organic synthesis

Organic Synthesis Search

Browse synthetic transformations by the desired bond formation

The graphical index, with various options and links to follow, should help in developing new ideas. Please try to search the site directly if you do not find your desired reaction.

Target-oriented synthesis (TOS) versus diversity-oriented synthesis (DOS)

TOS and DOS are two very different approaches to organic synthesis, with different aims and scopes. If a specific target molecule for example offers properties of interest - such as a natural compound with activity against a specific disease - a short, convenient and high-yielding synthesis should be developed, that allows a fast chemical approach to that desired molecule. In this case, a synthetic chemist takes a piece of paper or software and tries to split this large molecule into smaller molecules (disconnections) and repeats that procedure iteratively, until commercially available building blocks result. This careful analysis - which is known as retrosynthesis - might result in several options that lead to the desired molecule. In the forward synthesis starting from the commercially available building blocks, several synthetic steps are then needed until the desired molecule can be isolated. In a linear approach, unpredictable bottlenecks can cause an unwanted return to the synthetic starting point so enough intermediate is available, whereas a convergent approach limits the risk of a total failure, as only a few steps of a branch synthesis must be repeated. Depending on such risks, experienced organic chemists run the first few reactions at a larger scale and store all intermediates, so they can optimize a reaction step if needed.

A totally different approach is diversity-oriented synthesis. In DOS, organic chemists try to use versatile functional groups and let them react with a broad range of substrates, so the products cover a broad chemical space. That include new functionalities with totally different binding properties, ring sizes and 3D configurations. Such libraries, that are often generated in a combinatorial chemistry approach, can be screened against multiple properties (drug activities, physicochemical properties). Molecules that show properties of interest (for example a low IC 50 value for the binding affinity against a specific enzyme) then serve as lead compounds. Synthesizing smaller variations of that initial lead structure in parallel syntheses gives an idea about structure and activity relation (SAR) of specific modifications.

If one of these smaller variations results in a more active compound, this molecule becomes the new lead compound. Over the time, the syntheses also become more targeted, as people develop a better understanding of the SAR and focus on specific molecules rather than diversity.

Site Search any all words

Main Categories

• Organic Reactions
• Org. Chem. Highlights
• Abstracts ( Info )
• Presentations
• Chemistry Tools
• Ads & Imprint

Popular Subcategories

• Name Reactions
• Protecting Groups
• Organic Synthesis

• A Message from the Chair
• Latest News
• Calendar of Events
• Newsletter Archive
• Contact Us/Feedback
• Membership Benefits (PDF)
• Executive Committee
• Frequently Asked Questions
• History of the Division
• Awards Earned by the Division

• indicates you must be a member of the DOC to access.
• Eminent Organic Chemists’ Videos
• Historic Chemistry Videos
• Past Virtual Symposia
• NOS Lecture Videos
• Other Organic Chemistry Videos
• Organic Chemistry Journals

Organic Syntheses

• Organic Reactions
• Chemistry Reference Resolver
• Organic Related RSS Feeds
• Organic Chemistry Data Website
• Links to Organic Chemistry Sources
• Green Organic Chemistry
• Organic Solvent Data
• Synthetic Organic Groups (Organiclinks)
• Famous Organic Chemists
• ACS Technical Divisions
• National Organic Chemistry Symposium Website
• NOS History
• Historical Perspective of NOS in JOC
• National Meeting General Info
• ORGN Technical Program Abstracts
• Abstract Submission Guidelines
• Symposia Proposals
• Travel Awards
• Support for ACS Regional Meetings
• Graduate Research Symposium (GRS)
• DOC Virtual Symposium Series
• Empowering Women in Organic Chemistry (EWOC)
• Other Organic Chemistry Meetings of Interest
• Meeting Support
• ACS Office of Diversity, Equity, Inclusion, & Repsect
• Adams Award
• Gassman Award
• Leete Award
• Academic Young Investigator’s Symposium (YAS)
• Early Career Investigator Award
• Mid-Career Investigator Award
• Technical Achievement Awards (TAOC)
• Other Awards
• Summer Undergraduate Research Fellowships (SURF)
• Undergraduate Award
• History of the Graduate Research Fellowship Program
• Troyansky Endowment
• JOC Outstanding Publication of the Year Award Lectureship
• Org. Lett. Publication of the Year
• Organometallics Distinguished Author Award
• ACS Catalysis Lectureship Award
• OPR&D Lectureship Award

• Employment Resources
• ACS Career Resources
• Volunteer Opportunities
• Resources for Internships and Fellowships

Since 1921, Organic Syntheses has provided the chemistry community with annual collections of detailed, reliable, and carefully checked procedures for the synthesis of organic compounds. Some procedures describe practical methods for the preparation of specific compounds of interest, while other procedures illustrate important synthetic methods with general utility. Each procedure is written in considerably more detail as compared to typical experimental procedures in other journals, and each reaction and all characterization data has been carefully “checked” for reproducibility in the laboratory of a member of the Board of Editors. In order for an article to be accepted for publication, each reaction must be successfully repeated at least twice in an editor’s laboratory, with yields and selectivity similar to that reported by the submitting authors. Over 2500 articles have been published to date, incorporating procedures for conducting well over 5000 different synthetic reactions.

Organic Syntheses Inc. and the ACS Organic Division have worked closely together for a large number of years. In fact, Organic Syntheses funds the Roger Adams Award and until the program was discontinued, several of the Graduate Fellowships. In their place, they now provide funding to the DOC’s Graduate Research Symposium (GRS) , several Gordon Conferences, and they also have a summer grant program for faculty at principally undergraduate institutions (PUIs).

The Organic Syntheses open access website ( http://www.orgsyn.org/ ) allows users to access individual articles via graphical tables of contents and also to conduct keyword and substructure searches of the database of the reactions highlighted in the volumes via a free ChemDraw plugin.

Organic Syntheses regularly financially supports Organic Division programs and they cosponsor the Roger Adams Award which recognizes and encourage outstanding contributions to research in organic chemistry defined in its broadest sense. The award is presented biennially in odd-numbered years at the National Organic Chemistry Symposium (NOS) . A video of the 2023 Award presentation including some of the history of Organic Syntheses is available here .

In 2018, Organic Syntheses announced the free Org. Syn. Mobile App for iOS (iPhone and iPad) which can be downloaded for free from the App Store . An Android version of the app is under development although a date for its release is not yet available.

For more information about Organic Syntheses visit http://www.orgsyn.org/ .

Latest from OrgSynth

New article published: Nickel-catalyzed Amination of Aryl Chlorides [...]

New article published: Tandem Fragmentation/Olefination for Production of Neopentylene-Tethered 1,6-Enynes from Dimedone [...]

New article published: Synthesis of 5-(1-Diazo-2-ethoxy-2-oxoethyl) dibenzo[b,d]thiophenium Triflate [...]

New article published: Discussion Addendum for: Mild Conversion of Tertiary Amides to Aldehydes Using Cp2Zr(H)Cl (Schwartz’s Reagent) [...]

New article published: Preparation of a Radical Clocks Bearing Carbonyl Groups: Synthesis of N-Methoxy-N-methylspiro[cyclopropane-1,9'-fluorene]-2-carboxamide [...]

New article published: Cobalt-catalyzed Radical Olefin Isomerization: Retrocycloisomerization of (−)-Caryophyllene Oxide to (−)-Humulene Oxide II [...]

New article published: Metal-Free Aziridination of Alkenes with Free Amines by Thianthrenation [...]

New article published: Synthesis of Deuterated 1H-Indole-3-carboxaldehyde via Catalytic Vilsmeier-Haack Reaction [...]

New article published: Synthesis of Secondary Benzylic Alcohols by Reductive Arylation of Aldehydes: α-Phenyl-6-quinolinemethanol [...]

New article published: Large Scale Epoxide Opening by a Pendant Silanol [...]

Themed collection Organic Synthesis

Α-borylalkyl radicals: their distinctive reactivity in modern organic synthesis.

In this review, we emphasise the importance of the generation of α-boryl carbon-centred radicals and their utilisation in synthesis.

Gold and hypervalent iodine( III ): liaisons over a decade for electrophilic functional group transfer reactions

Building on mechanistic perspective, the review intends to demonstrate how the uniqueness of Au-catalysts has realized a myriad of electrophilic functional group transfer reactions with the use of hypervalent iodine( III ) reagents over the last decade.

Recent advances in metal-catalysed asymmetric sequential double hydrofunctionalization of alkynes

Recent advances in various metal-catalysed asymmetric sequential double hydrofunctionalizations of alkynes have been highlighted in this feature article.

Twenty-five years of bis-pentafluorophenyl borane: a versatile reagent for catalyst and materials synthesis

Highlights of the extensive chemistry and applications of bis-pentafluorophenyl borane (“Piers’ borane”) from the 25 years since its first appearance are featured.

Photoinduced deaminative strategies: Katritzky salts as alkyl radical precursors

Primary amines are one of the most predominant functional groups found in organic molecules. This review covers the most recent developments on photocatalytic deaminative strategies by using Katritzky Salts as alkyl radical reservoirs.

Transition metal-catalyzed sp 3 C–H activation and intramolecular C–N coupling to construct nitrogen heterocyclic scaffolds

Nitrogen heterocycles are of great medicinal importance, and the construction of nitrogen heterocyclic scaffolds has been one of the focuses in synthetic organic chemistry.

Enantioselective synthesis of multi-nitrogen-containing heterocycles using azoalkenes as key intermediates

The recently developed annulation reactions using azoalkenes as key intermediates show their great ability to construct diverse types of multi-nitrogen-containing heterocycles. In this feature article, we critically analysed the strategic development and the efficient transformation of azoalkenes to chiral heterocycles and α-functionalized ketone derivatives since 2010.

Continuous flow chemistry: where are we now? Recent applications, challenges and limitations

A general outlook of the changing face of chemical synthesis is provided in this article through recent applications of continuous flow processing in both industry and academia.

Copper-catalysed ortho -selective C–H bond functionalization of phenols and naphthols with α-aryl-α-diazoesters

An unprecedented CuCl 2 -catalysed chemo- and ortho -selective C–H bond functionalization of phenols and naphthols with diazoesters has been developed.

Combination of organocatalytic oxidation of alcohols and organolithium chemistry (RLi) in aqueous media, at room temperature and under aerobic conditions

Organocatalysis and highly-polar s-block organometallic chemistry (RLi) work together in water, under air and at room temperature for the selective and ultrafast synthesis of tertiary alcohols.

Regioselective biocatalytic self-sufficient Tishchenko-type reaction via formal intramolecular hydride transfer

Alcohol dehydrogenases catalyze the regioselective lactonization of dialdehydes via a bio-Tishchenko-like reaction. The nicotinamide-dependent self-sufficient reduction–oxidation sequence proceeds through a formal intramolecular hydride shift.

Highly regioselective ring-opening of epoxides with amines: a metal- and solvent-free protocol for the synthesis of β-amino alcohols

We herein report a metal- and solvent-free acetic acid-mediated ring-opening reaction of epoxides with amines.

Copper-catalyzed asymmetric silyl addition to alkenyl-substituted N -heteroarenes

Asymmetric conjugate addition of PhMe 2 SiBPin to a wide range of N -heteroaryl alkenes proceeded in the presence of a copper catalyst coordinated with a chiral phosphoramidite ligand to afford useful β-silyl N -heteroarenes in high yields and ees.

Copper mediated C(sp 2 )–H amination and hydroxylation of phosphinamides

Copper mediated C(sp 2 )–H amination and hydroxylation of arylphosphinic acid are accomplished by adopting phosphinamide as the directing group.

Synthesis of imides via palladium-catalyzed three-component coupling of aryl halides, isocyanides and carboxylic acids

A palladium-catalyzed, three-component synthesis of imides from feedstock aryl halides, carboxylic acids and isocyanides through the intermediacy of isoimide has been developed.

Palladium-catalyzed regioselective C–H alkynylation of indoles with haloalkynes: access to functionalized 7-alkynylindoles

A palladium-catalyzed exclusively selective alkynylation of indoles has been reported, affording concise access to 7-alkynylindoles from readily available starting materials.

Highly stereoselective intramolecular Buchner reaction of diazoacetamides catalyzed by a Ru( II )–Pheox complex

This work reports the first efficient enantioselective intramolecular Buchner reaction of diazoacetamides.

Direct bromocarboxylation of arynes using allyl bromides and carbon dioxide

An unprecedented multicomponent reaction involving arynes, allyl bromides, and CO 2 has been developed to construct various allyl o -bromobenzoate scaffolds.

A ruthenium-catalyzed free amine directed (5+1) annulation of anilines with olefins: diverse synthesis of phenanthridine derivatives

A ruthenium( II )-catalyzed cross-ring (5+1) annulation between 2-aminobiphenyls and activated olefins is disclosed for succinct synthesis of valuable phenanthridine scaffolds.

Pd-Catalyzed decarboxylative cross-coupling reactions of epoxides with α,β-unsaturated carboxylic acids

A Pd-catalyzed decarboxylative cross-coupling of α,β-unsaturated carboxylic acids with cyclic and acyclic epoxides has been developed.

Cu-Catalyzed highly selective reductive functionalization of 1,3-diene using H 2 O as a stoichiometric hydrogen atom donor

A novel high regio- and diastereo-selectivity reductive functionalization of 1,3-diene has been developed using H 2 O as a stoichiometric hydrogen atom donor.

A halogen-bonding-catalysed Nazarov cyclisation reaction

Various neutral, mono- and dicationic halogen bond donors were screened for their ability to act as catalysts in a Nazarov cyclisation reaction.

Highly active dinuclear cobalt complexes for solvent-free cycloaddition of CO 2 to epoxides at ambient pressure

Dinuclear Co-based catalysts are used for the coupling reaction of epoxides and CO 2 in the presence of a cocatalyst.

Pd-catalyzed synthesis of α,β-unsaturated ketones by carbonylation of vinyl triflates and nonaflates

A general and highly chemoselective Pd-catalyzed protocol for the synthesis of α,β-unsaturated ketones by carbonylation of vinyl triflates and nonaflates is presented. Applying the specific monophosphine ligand cataCXium® A , the synthesis of various vinyl ketones as well as carbonylated natural product derivatives proceeds in good yields.

Visible light mediated, metal-free carbene transfer reactions of diazoalkanes with propargylic alcohols

The photolysis of donor–acceptor diazoalkanes in the presence of propargylic alcohols furnishes valuable, sterically demanding tetra-substituted cyclopropenes in high yield under metal-free conditions.

Synthesis of polysubstituted 3-aminoindenes via rhodium-catalysed [3+2] cascade annulations of benzimidates with alkenes

A novel Rh-catalysed intermolecular [3+2] cascade cyclization of benzimidates and alkenes has been developed to assemble polysubstituted 3-aminoindenes, which exhibits good functional-group tolerance and excellent regioselectivity.

Fe( II )-Catalyzed alkenylation of benzylic C–H bonds with diazo compounds

A direct alkenylation of benzylic C(sp 3 )–H bonds with diazo compounds with FeCl 2 as the catalyst and DDQ as the oxidant has been developed.

Synthesis of dihydroquinolinones via iridium-catalyzed cascade C–H amidation and intramolecular aza-Michael addition

An iridium-catalyzed annulation of chalcones with sulfonyl azides via cascade C–H amidation and aza-Michael addition is developed to provide 2-aryl-2,3-dihydro-4-quinolones.

Palladium-catalyzed oxidative borylation of conjugated enynones through carbene migratory insertion: synthesis of furyl-substituted alkenylboronates

A new method for the synthesis of furyl-substituted alkenylboronates has been developed by palladium-catalyzed oxidative borylation reaction of conjugated enynones.

Palladium-catalyzed olefination of aryl/alkyl halides with trimethylsilyldiazomethane via carbene migratory insertion

One-pot formation of ( E )-vinyl silanes and ( E )-silyl-substituted α, β-unsaturated amides could be completed easily via palladium carbene migratory insertion in good yields and high chemoselectivity.

An efficient method for retro -Claisen-type C–C bond cleavage of diketones with tropylium catalyst

We report a new convenient and efficient method utilizing the tropylium ion as a mild and environmentally friendly organocatalyst to mediate retro -Claisen-type reactions.

Mechanistic and asymmetric investigations of the Au-catalysed cross-coupling between aryldiazonium salts and arylboronic acids using (P,N) gold complexes

Aryldiazonium salts and arylboronic acids were coupled via three different pathways from (P,N)–AuCl complexes, with enantiomeric excesses up to 26%.

Arylation of benzyl amines with aromatic nitriles

The C(sp 3 )–H arylation of benzyl amines with aromatic nitriles for the synthesis of diarylmethylamines was realized without the assistance of transition-metal and photoirradiation.

C–H functionalisation of aldehydes using light generated, non-stabilised diazo compounds in flow

Here we explore further the use of oxadiazolines, non-stabilised diazo precursors which are bench stable, in direct, non-catalytic, aldehyde C–H functionalisation reactions under UV photolysis in flow and free from additives.

Reinventing the De Mayo reaction: synthesis of 1,5-diketones or 1,5-ketoesters via visible light [2+2] cycloaddition of β-diketones or β-ketoesters with styrenes

Taking a different route yields the same product of the De Mayo reaction, but allows the use of visible light instead of UV irradiation.

Visible-light photocatalytic bicyclization of β-alkynyl propenones for accessing diastereoenriched syn -fluoren-9-ones

A novel visible-light photocatalytic bicyclization of β-alkynyl propenones with α-bromocarbonyls for highly diastereoselective synthesis of richly decorated syn -fluoren-9-ones is described.

Cyanomethyl anion transfer reagents for diastereoselective Corey–Chaykovsky cyclopropanation reactions

A readily available sulfonium salt opens up new synthetic pathways to access nitrile cyclopropanes in a highly diastereoselective fashion.

Gold-catalyzed annulations of N -aryl ynamides with benzisoxazoles to construct 6 H -indolo[2,3- b ]quinoline cores

This work reports new annulations of N -aryl ynamides with benzisoxazoles to form 6 H -indolo[2,3- b ]quinoline derivatives.

Asymmetric synthesis of polysubstituted methylenecyclobutanes via catalytic [2+2] cycloaddition reactions of N -allenamides

Enantioselective acyl-transfer catalysis by fluoride ions

The asymmetric nucleophilic catalysis by fluoride ions at a carbon-based electrophile has been demonstrated for the first time.

Reductive cyclisations of amidines involving aminal radicals

The first general study of aminal radical cyclisations, triggered by reduction of amidines with SmI 2 , delivers quinazolinones with complete diastereocontrol.

Trisubstituted olefin synthesis via Ni-catalyzed hydroalkylation of internal alkynes with non-activated alkyl halides

The stereoselective synthesis of tri-substituted alkenes is challenging.

Synthesis of bench-stable solid triorganoindium reagents and reactivity in palladium-catalyzed cross-coupling reactions

Triorganoindium reagents can be isolated as bench-stable solid R 3 In(DMAP) complexes and show excellent reactivity in palladium-catalyzed cross-coupling reactions.

Palladium-catalyzed carbene/alkyne metathesis with enynones as carbene precursors: synthesis of fused polyheterocycles

An unprecedented palladium-catalyzed carbene/alkyne metathesis reaction of alkyne-tethered enynones is described, which delivers fused-furans in moderate to good yields.

About this collection

Organic synthesis is far from the level that many people assume. Progress is continuing, but there will not be any dramatic developments. It is more like a glacier that gradually moves forward until it is has finally covered an entire region, but it will still be centuries before synthesis has acquired the status that many people already ascribe to it today.” G. Stork in Nachr. Chem. Techn. 1972, 20, 147. Collated by Antonio Echavarren (ICIQ, Spain), this collection reflects the continuous efforts at gradually moving forward in synthetic methodology.

Advertisements

• school Campus Bookshelves
• menu_book Bookshelves
• perm_media Learning Objects
• login Login
• how_to_reg Request Instructor Account
• hub Instructor Commons

Margin Size

• Download Page (PDF)
• Download Full Book (PDF)
• Periodic Table
• Physics Constants
• Scientific Calculator
• Reference & Cite
• Tools expand_more
• Readability

selected template will load here

This action is not available.

Chapter 18: Organic Synthesis

• Last updated
• Save as PDF
• Page ID 23619

\( \newcommand{\vecs}[1]{\overset { \scriptstyle \rightharpoonup} {\mathbf{#1}} } \)

\( \newcommand{\vecd}[1]{\overset{-\!-\!\rightharpoonup}{\vphantom{a}\smash {#1}}} \)

\( \newcommand{\id}{\mathrm{id}}\) \( \newcommand{\Span}{\mathrm{span}}\)

( \newcommand{\kernel}{\mathrm{null}\,}\) \( \newcommand{\range}{\mathrm{range}\,}\)

\( \newcommand{\RealPart}{\mathrm{Re}}\) \( \newcommand{\ImaginaryPart}{\mathrm{Im}}\)

\( \newcommand{\Argument}{\mathrm{Arg}}\) \( \newcommand{\norm}[1]{\| #1 \|}\)

\( \newcommand{\inner}[2]{\langle #1, #2 \rangle}\)

\( \newcommand{\Span}{\mathrm{span}}\)

\( \newcommand{\id}{\mathrm{id}}\)

\( \newcommand{\kernel}{\mathrm{null}\,}\)

\( \newcommand{\range}{\mathrm{range}\,}\)

\( \newcommand{\RealPart}{\mathrm{Re}}\)

\( \newcommand{\ImaginaryPart}{\mathrm{Im}}\)

\( \newcommand{\Argument}{\mathrm{Arg}}\)

\( \newcommand{\norm}[1]{\| #1 \|}\)

\( \newcommand{\Span}{\mathrm{span}}\) \( \newcommand{\AA}{\unicode[.8,0]{x212B}}\)

\( \newcommand{\vectorA}[1]{\vec{#1}}      % arrow\)

\( \newcommand{\vectorAt}[1]{\vec{\text{#1}}}      % arrow\)

\( \newcommand{\vectorB}[1]{\overset { \scriptstyle \rightharpoonup} {\mathbf{#1}} } \)

\( \newcommand{\vectorC}[1]{\textbf{#1}} \)

\( \newcommand{\vectorD}[1]{\overrightarrow{#1}} \)

\( \newcommand{\vectorDt}[1]{\overrightarrow{\text{#1}}} \)

\( \newcommand{\vectE}[1]{\overset{-\!-\!\rightharpoonup}{\vphantom{a}\smash{\mathbf {#1}}}} \)

‘There could be ART in Organic Synthesis’ declared the inimitable monarch of organic synthesis, Professor R.B. Woodward. His school unveiled several elegant approaches covering a variety of complex structures and broke new grounds to define the art of organic synthesis. ‘If organic synthesis is a branch of science, what is the LOGIC of organic synthesis?’ marveled several others. The development of the concept of logical approaches towards synthesis has been evolving over the past several decades. A few stalwarts focused their attention on this theme and attempted to evolve a pattern to define this logic. There is no doubt that all of us who dabble with synthesis contribute our small bit in the magnificent direction. A few names stand out in our minds for their outstanding contributions. Notable contributions came from the schools of J.A. Marshal, E.J. Wenkert, G. Stock, S Hanessian, E.E. van Tamalen, S. Masamune, R.B. Woodward, E.J. Corey and several others. More focused on this theme were the contributions from the school of E.J. Corey.

The period 1960 – 1990 witnessed the evolution of this thought and the concept bloomed into a full-fledged topic that now merits a separate space in college curriculum. Earlier developments focused on the idea of ANTITHETIC APPROACHES and perfected the art of DISCONNECTION via RETROSYNTHESIS. This led to logical approaches for the construction of SYNTHETIC TREES that summarized various possible approaches for the proposed Target structure. All disconnections may not lead to good routes for synthesis. Once the synthetic tree was constructed, the individual branches were analyzed critically. The reactions involved were looked into, to study their feasibility in the laboratory, their mechanistic pathways were analyzed to understand the conformational and stereochemical implications on the outcome of each step involved and the time / cost factors of the proposed routes were also estimated. The possible areas of pitfall were identified and the literature was critically scanned to make sure that the steps contemplated were already known or feasible on the basis of known chemistry. In some cases, model compounds were first constructed to study the feasibility of the particular reaction, before embarking on the synthesis of the complex molecular architecture. Thus a long process of logical planning is now put in place before the start of the actual synthetic project. In spite of all these careful and lengthy preparations, an experienced chemist is still weary of the Damocles Sword of synthesis viz., the likely failure of a critical step in the proposed route(s), resulting in total failure of the entire project. All achievements are 10% inspiration and 90% perspiration. For these brave molecular engineers, sometimes also called chemists, these long-drawn programs and possible perils of failures are still worth, for the perspiration is enough reward.

A sound knowledge of mechanistic organic chemistry, detailed information on the art and science of functional group transformations, bond formation and cleavage reactions, mastery over separation and purification techniques and a sound knowledge of spectroscopic analysis are all essential basics for the synthesis of molecules. A synthetic chemist should also be aware of developments in synthetic strategies generated over the years for different groups of compounds, which include Rules and guidelines governing synthesis. Since organic chemistry has a strong impact on the development of other sister disciplines like pharmacy, biochemistry and material science, an ability to understand one or more of these areas and interact with them using their terminologies is also an added virtue for a synthetic chemist. With achievements from synthesis of strained molecules (once considered difficult (if not impossible) to synthesize, to the synthesis of complex, highly functionalized and unstable molecules, an organic chemist could now confidently say that he could synthesize any molecule that is theoretically feasible. This is the current status of the power of organic synthesis. Based on the task assigned to the chemist, he would select a Target molecule for investigation and devise suitable routes for synthesis.

Protection and Deprotection Strategies in Organic syntheses

For the manipulation of functional groups and formation of new covalent bonds we make use of a large number of Reagents and Name Reactions. In complex organic syntheses, the starting materials and intermediates in the synthetic scheme often have more than one reactive functional group. A few such multifunctional building blocks are shown below to illustrate this point (Fig 4.1.1 ) . While working on such complex

molecules, it is often necessary to protect some groups to enable selective working at the desired locations only. Organic chemists have heavily relied on such protection / deprotection strategies and have diligently developed protecting (masking) and deprotection (unmasking) protocols. We would discuss some of the important protecting groups in this chapter.

Before proceed further, it must be emphasized here that this protocol should be applied only after alternate options have been critically analyzed. This is because protection / deprotection strategy involves an increase of at least two more critical steps, adding to the length of the synthesis and consequent drop in overall yields of the desired compound. In large-scale reactions, this leads to a huge impact to the Atom Economy and pollution cost of the synthetic process. All this translates into an increase in the overall cost of the final drug molecule.

Protection Strategies

Group / Site Selective Reagent: Protection / deprotection is not always required whenever you see a multiplicity of functional groups. You could solve the selectivity issue by using site selective reactions / reagents. By choosing an appropriate selective reagent to suit the scheme on hand, you could selectively attack only one of the reactive sites. Consider an olefinic ketone (Fig 4.1.2) . Sodium borohydride reduction in methanol as solvent could selectively reduce the keto- group to a secondary alcohol

leaving the olefin undisturbed. On the other hand, diborane reagent in THF as solvent would be a reagent of choice when the selective reduction at the olefin moiety is desired. Diborane reduction of an olefin is several times faster than reduction of ketones. The oxidative cleavage of borane product is also selective. Thus, you can avoid the protection / deprotection strategy by employing a selective reagent. In C – C bond formation reactions we come across several such site-selective reagents. One such reagent widely used in research is the Wittig reagent. They attack the aldehyde or ketone selectively in the presence of ester, nitrile. olefin etc..

Selective Protection

In the case of a molecule like 4.1.3A (Fig 4.1.3) bearing an olefin and a carboxylic acid, the –COOH group is several times more reactive than the olefin towards diborane reduction.

Hydroboration / oxidation reduces the acid to a primary alcohol, leaving the olefin unaffected. On the other hand, if you need a selective reduction of olefin, the acid group has to be processed through a selective protection / deprotection sequence as shown in (Fig 4.1.3)

Compound 4.1.4A illustrates several important points in Protection / Deprotection protocol. Both the functional groups could react with a Grignard Reagent. Carboxylic acid group would first react with one mole of the Grignard Reagent to give a carboxylate anion salt. This anion does not react any further with the reagent. When two moles of Grignard Reagent are added to the reaction mixture, the second mole attacks the ketone to give a tertiary alcohol. On aqueous work-up, the acid group is regenerated. Thus, the first mole of the reagent provides a selective transient protection for the –COOH group. Once the acid group is esterified, such selectivity towards this reagent is lost. The reagent attacks at both sites. If reaction is desired only at the ester site, the keto- group should be selectively protected as an acetal. In the next step, the grignard reaction is carried out. Now the reagent has only one group available for reaction. On treatment with acid, the ketal protection in the intermediate compound is also hydrolyzed to regenerated the keto- group.

Orthogonal Protection or Differential Protection

Orthogonal protection is a strategy that allows deprotection of multiple protective groups one at a time, each with a dedicated set of reagents and / reaction conditions without affecting the other. This technique is best illustrated with peptide bond formation and associated deprotection reactions. An amino acid has two functional groups –N H 2 and –COOH. When two amino acids (A and B) react under conditions for the peptide bond condensation reaction, a mixture of 4 dipeptides (at least) could be formed as shown below.

\[ \ce{ A + B \rightarrow A-A + A-B + B-A + B-B}\]

If we are interested in only one product A – B, we have to do selective protections and selective deprotections in a proper sequence. Consider the following peptide bond formation reaction.

In order to get only one product A – B, we should protect the N – terminal of ‘A’ and C – terminal of ‘B’. Let us look closely at two different dipeptide formation schemes. In the following sequence, the C – terminal is protected in two different ways for one amino acid. For the second amino acid, the N – terminal is protected with an acid labile Boc- protection.

In the next step, the two monoprotected amino acids are coupled as shown below.

Take a close look at both the products. In the first product, both protections are acid sensitive. If the final product desired is the protection-free dipeptide, this is indeed a short route.

If the desired product is a mono-protected dipeptide, then selective deprotection is the preferred reaction. This is feasible only when we use starting compounds that are differentially protected. This is called Orthogonal Protection .

Similar techniques are available for other functional groups as well. Let us now learn more on Protection / Deprotection for some important functional groups.

Protection of R – COOH Group

In the introduction, we have seen that carboxylate ion lends protection to an attack of Grignard reagents at this carbonyl carbon. However, this is not sufficient for a vast variety of reagents. Meyer’s 2-oxazolines mask an acid function while activating the α- position for lithiation reaction. The use of this group as protection for –COOH group is rare.

Protection of Aldehydes and Ketones

Since alcohols, aldehydes and ketones are the most frequently manipulated functional groups in organic synthesis, a great deal of work has appeared in their protection / deprotection strategies. In this discussion let us focus on the classes of protecting groups rather than an exhaustive treatment of all the protections.

There are two general methods for the introduction of this protection. Transketalation is the method of choice when acetals (ketals) with methanol are desired. Acetone is the by-product, which has to be removed to shift the equilibrium to the right hand side. This is achieved by refluxing with a large excess of the acetonide reagent. Acetone formed is constantly distilled. In the case of cyclic diols, the water formed is continuously removed using a Dean-Stork condenser (Fig 4.1.6) .

The rate of formation of ketals from ketones and 1,2-ethanediol (ethylene glycol), 1,3-propanediol and 2,2-dimethyl-1,3-propanediol are different. So is the deketalation reaction. This has enabled chemists to selectively work at one center. The following examples from steroid chemistry illustrate these points (Fig 4.1.7) .

The demand for Green Chemistry processes has prompted search for new green procedures. Some examples from recent literature are given here (Fig 4.1.8) .

Compared with their oxygen analogues, thioketals markedly differ in their chemistry. The formation as well as deprotection is promoted by suitable Lewis acids. The thioacetals are markedly stable under deketalation conditions, thus paving way for selective operations at two different centers. When conjugated ketones are involved, the ketal formation (as well as deprotection) proceeds with double bond migration. On the other hand, thioketals are formed and deketalated without double bond migration (Fig 4.1.9) .

Protection of Amino groups (-NH2 &–NH )

N-Acetyl (N – COC H 3 ), N – Benzoyl (N – COPh) Protections

These are the classical protecting groups for primary and secondary amines. The reagents are cheap and the protocol is simple. Such amides generally need drastic conditions for deprotection, though the yields are generally good (Fig 4.1.10) . A standard procedure is refluxing in aqueous alkali or aqueous mineral acid. Due to the drastic conditions, care should be exercised in this procedure to ensure racemi zation is avoided. Amides are generally crystalline solids that are easily purified by crystallization. When the protection is introduced at the early stages of a long synthetic scheme and a very stable protection is desired (as in nucleotide synthesis) an amide is the most preferred protection.

Several more labile amide bonds have been investigated. The amides of trifluoroacetic acid are of special interest. The introduction as well as cleavage is simple and mild {Fig 4.1.11) .

A recent report in amide hydrolysis is given below.

N – Phthaloyl Protection (N – Pht )

Mechanism for NaB H 4 Reduction of N – Pht

N – Carboxylic acid Esters as protective groups

As described above, the amide bonds are very strong. On the other hand, the ester bonds are easily cleaved by mild base conditions. A carboethoxy protection on amine has an amide bond as well as an ester bond. Since N – COOH groups obtained on hydrolysis are very unstable, this protection provides a large family of protective groups for primary and secondary amines.

N – Carboethoxycarbonly (N – COOEt) and Carbobenzyloxycarbonyl (N – COOCH2Ph) (N – Cbz or N – Z) Protections:

These groups are easily introduced using the corresponding chloroformate esters. Anhydrides or mixed anhydrides under mild basic conditions. Both these protections could be removed under prolonged stirring with base at room temperature. Though mild, some racemisation is sometimes observed. The N – Cbz protection has an added advantage in that it could be easily cleaved under hydrogenolysis conditions (Fig 4.1.14) . N – Cbz Protection is however stable to acidic conditions. Compare this with –Boc protection discussed below.

Tert-Butyloxycarbonyl Protection ( N – COOBut, N – Boc)

The Tert-Butyloxycarbonyl Protection could be introduced and removed under very mild acid conditions. This protection is stable to alkali and hydrogenolysis (Fig 4.1.15) . Thus, N – Z and N – Boc are complimentary as protective groups.

N – Fluoromethyleneoxycarbonyl Protection (Fmoc)

This UV active protecting group is very popular in Solid Phase Peptide Synthesis (SPPS) protocols. Protection as well as deprotection steps proceed under mild conditions in good yields (Fig 4.1.16) .

The mechanism for Fmoc deprotection is shown in (Fig 4.1.17)

N – Silylation

Silylation is a common protection for active hydrogen on heteroatoms. In the case of N – Si bond, quaternary ammonium fluorides cleave this bond (Fig 4.1.18) .

N – Tosylation (N – Tos)

This protection is very stable. N – Tosylation is easily carried out through acid chloride procedure. It is cleaved by solvated electron cleavage reaction. When this group is attached to a primary amine, the –NH group becomes very acidic (Fig 4.1.19) .

Protection of – OH Groups

Acetates ( – Ac) and benzoates (– Obz)):

The – OH group protection chemistry has been extensively investigated. The classical protection is the formation of esters of aliphatic and aromatic carboxylic acids. Aromatic esters are comparatively difficult to hydrolyze under mild base condition. This provides an opportunity for selective deprotection protocols (Fig 4.1.20) . Note that this protection is sensitive to acid as well as base conditions.

Methyl (– OMe) and Benzyl (R – OBn) Ethers

An ether group is one amongst the most stable functional groups. Hence, this group has been the most favored protecting group. Deprotection was a problem. In the early part of the twentieth century, the only procedure was refluxing with aqueous HI or HBr. In recent years several new procedures have appeared for effective removal under mild conditions. The special feature of the benzyl ethers is that this protection is readily removed under neutral hydrogenolysis conditions (Fig 4.1.21) . Substituents like – OMe or – NO2 could be introduced on the benzene ring to modify the reactivity at the protection site.

When an olefin could compete at the hydrogenolysis procedure, the following sequence appears to be an alternate procedure (Fig 4.1.22) .

Allyl ether is a recent introduction in – OH protection. The versatility of this protection could be seen in the following examples (Fig 4.1.23) .

Silyl Ethers (R – OSi R 3 )

The oxygen – silicon sigma bond is stable to lithium and Grignard reagents, nucleophiles and hydride reagents but very unstable to water and mild aqueous acid and base conditions. A silyl ether of secondary alcohol is less reactive than that of a primary alcohol. The O – trimethylsilyl (O – SiMe3) was first protection of this class. (Fig 4.1.24) .

Replacement of methyl group with other alkyl and aryl groups gives a large variety of silyl ether with varying degrees of stability towards hydrolysis (Fig 4.1.25) .

The following examples illustrate the selectivity in formation and hydrolysis of this group (Fig 4.1.26) .

Tetrahydropyranyl ether (– OTHP) and Tetrahydrofuranyl ether (– OTHF)

These protective groups for alcohols are in fact acetals. They are synthesized using the dihydropyran (DHP) and dihydrofuran (DHF) respectively. They behave like acetals in their stability and cleavage (Fig 4.1.27) . The rate of formation and cleavage for these two groups differ, which finds application for differential protection of alcohols.

These protective groups found extensive use in synthesis. However, two major drawbacks were soon observed.

• A new stereopoint is generated while introducing this protection. Though it is not relevant from the point of view of the target molecule, in chiral molecules this created diastereomer problems in spectroscopy (NMR and MS) and chromatography.
• These ethers occasionally caused explosions in hydroboration procedures due to peroxide formation. The diastereomer problem was solved by the introduction of O-methyleneoxymethyl ether ( – O – MOM) and O – methyleneoxybenzyl ether (R – O – MOB) (Fig 4.1.28) . Several other modifications are now available.

Protection of vic – Diols

On reaction with benzaldehyde or acetone with suitable acid catalyst, vic-diols form cyclic acetals. This in fact is a proof for the existence of vic-diols in the molecule. They are acetal protections and therefore behave as acetals in their chemistry (Fig 4.1.29)

The above discussions are just a glimpse of the vast literature on this topic. When more than one competing functional groups are present in a molecule, it may be necessary to introduce at least one protection and one deprotection step in the synthetic scheme. This adds not only to the length of the synthetic scheme, but also to the cost of the final compound. With growing awareness in Green Chemistry , chemists have been trying to reduce this protocol to a minimum or preferably avoid this altogether. Several protection free syntheses of natural products are known in the literature. We would discuss this topic at the end of this chapter.

Further Reading

• Greene T. W., Protective Groups in Organic Synthesis. Wiley. N. Y., (1980), (1991).
• Smith M. B., Organic Synthesis, McGraw-Hill Inc, N. Y., (1994).
• Djerassi C., Steroid Reactions –An Outline for Organic Chemists, Holden-Day nc. San Francisco )1963).
• Advanced Organic Chemistry: Principles Tools and Logic of Synthesis, R. Balaji Rao, Vishal Publishing Co., Jalandhar, India (2012).
• Amino Acids, Peptides and Proteins in Organic Chemistry, Vol 4, Ed by Andrew B. Hughes (2011) Wiley-VCH; Protection Reactions, V.V. Sureshbabu and N. Narendra page 1 – 97.

4.2 Disconnection of bonds

Having chosen the TARGET molecule for synthesis, the next exercise is to draw out synthetic plans that would summarize all reasonable routes for its synthesis. During the past few decades, chemists have been working on a process called RETROSYNTHESIS. Retrosynthesis could be described as a logical Disconnection at strategic bonds in such a way that the process would progressively lead to easily available starting material(s) through several synthetic plans. Each plan thus evolved, describes a ‘ROUTE’ based on a retrosynthesis. Each disconnection leads to a simplified structure. The logic of such disconnections forms the basis for the retroanalysis of a given target molecule. Natural products have provided chemists with a large variety of structures, having complex functionalities and stereochemistry. This area has provided several challenging targets for development of these concepts. The underlining principle in devising logical approaches for synthetic routes is very much akin to the following simple problem. Let us have a look of the following big block, which is made by assembling several small blocks (Fig 4.2.1) . You could easily see that the large block could be broken down in different ways and then reassembled to give the same original block.

Now let us try and extend the same approach for the synthesis of a simple molecule. Let us look into three possible ‘disconnections’ for a cyclohexane ring as shown in Figure 4.2.2.

In the above analysis we have attempted to develop three ways of disconnecting the six membered ring. Have we thus created three pathways for the synthesis of cyclohexane ring? Do such disconnections make chemical sense? The background of an organic chemist should enable him to read the process as a chemical reaction in the reverse (or ‘retro-‘) direction. The dots in the above structures could represent a carbonium ion, a carbanion, a free radical or a more complex reaction (such as a pericyclic reaction or a rearrangement). Applying such chemical thinking could open up several plausible reactions. Let us look into path b, which resulted from cleavage of one sigma bond. An anionic cyclisation route alone exposes several candidates as suitable intermediates for the formation of this linkage. The above analysis describes only three paths out of the large number of alternate cleavage routes that are available. An extended analysis shown below indicates more such possibilities (Fig 4.2.3) . Each such intermediate could be subjected to further disconnection process and the process continued until we reach a reasonably small, easily available starting materials. Thus, a complete ‘SYNTHETIC TREE’ could be constructed that would summarize all possible routes for the given target molecule.

4.3 Efficiency of a route

A route is said to be efficient when the ‘overall yield’ of the total process is the best amongst all routes investigated. This would depend not only on the number of steps involved in the synthesis, but also on the type of strategy followed. The strategy could involve a ‘linear syntheses’ involving only consequential steps or a ‘convergent syntheses’ involving fewer consequential steps. Figure 4.3.1 shown below depicts a few patterns that could be recognized in such synthetic trees. When each disconnection process leads to only one feasible intermediate and the process proceeds in this fashion

all the way to one set of starting materials (SM), the process is called a Linear Synthesis . On the other hand, when an intermediate could be disconnected in two or more ways leading to different intermediates, branching occurs in the plan. The processes could be continued all the way to SMs. In such routes different branches of the synthetic pathways converge towards an intermediate. Such schemes are called Convergent Syntheses .

The flow charts shown below (Fig 4.3.2) depicts a hypothetical 5-step synthesis by the above two strategies. Assuming a very good yield (90%) at each step (this is rarely seen in real projects), a linier synthesis gives 59% overall yield, whereas a convergent synthesis gives 73% overall yield for the same number of steps..

4.4 Problem of substituents and stereoisomers

The situation becomes more complex when you consider the possibility of unwanted isomers generated at different steps of the synthesis. The overall yield drops down considerably for the synthesis of the right isomer. Reactions that yield single isomers (Diastereospecific reactions) in good yields are therefore preferred. Some reactions like the Diels Alder Reaction generate several stereopoints (points at which stereoisomers are generated) simultaneously in one step in a highly predictable manner. Such reactions are highly valued in planning synthetic strategies because several desirable structural features are introduced in one step. Where one pure enantiomer is the target, the situation is again complex. A pure compound in the final step could still have 50% unwanted enantiomer, thus leading to a drastic drop in the efficiency of the route. In such cases, it is desirable to separate the optical isomers as early in the route as possible, along the synthetic route. This is the main merit of the Chiron Approach, in which the right starting material is chosen from an easily available, cheap ‘chiral pool’. We would discuss this aspect after we have understood the logic of planning syntheses. Given these parameters, you could now decide on the most efficient route for any given target.

Molecules of interest are often more complex than the plain cyclohexane ring discussed above. They may have substituents and functional groups at specified points and even specific stereochemical points. Construction of a synthetic tree should ideally accommodate all these parameters to give efficient routes. Let us look into a slightly more complex example shown in Figure 4.4.1 . The ketone 4.4.1A is required as an intermediate in a synthesis. Unlike the plain cyclohexane discussed above, the substitution pattern and the keto- group in this molecule impose some restrictions on disconnection processes.

Cleavage a: This route implies attack of an anion of methylisopropylketone on a bromo-component. Cleavage b: This route implies simple regiospecific methylation of a larger ketone that bears all remaining structural elements. Cleavage c: This route implies three different possibilities. Route C-1 envisages an acylonium unit, which could come from an acid halide or an ester. Route C-2 implies an umpolung reaction at the acyl unit. Route C-3 suggests an oxidation of a secondary alcohol, which could be obtained through a Grignard-type reaction. Cleavage d: This implies a Micheal addition.

Each of these routes could be further developed backwards to complete the synthetic tree. These are just a few plausible routes to illustrate an important point that the details on the structure would restrict the possible cleavages to some strategic points. Notable contributions towards planning organic syntheses came from E.J. Corey’s school. These developments have been compiles by Corey in a book by the title LOgIC OF CHEMICAL SYNTHESIS. These and several related presentations on this topic should be taken as guidelines. They are devised after analyzing most of the known approaches published in the literature and identifying a pattern in the logic. They need not restrict the scope for new possibilities. Some of the important strategies are outlined below.

4.5 Preliminary scan

When a synthetic chemist looks at the given Target, he should first ponder on some preliminary steps to simplify the problem on hand. Is the molecule polymeric? See whether the whole molecule could be split into monomeric units, which could be coupled by a known reaction. This is easily seen in the case of peptides, nucleotides and organic polymers. This could also be true to other natural products. In molecules like C-Toxiferin 1 (4.5.1A) (Fig 4.5.1) , the point of dimerisation is obvious. In several other cases, a deeper insight is required to identify the monomeric units, as is the case with Usnic acid (4.5.1B) . In the case of the macrolide antibiotic Nonactin (4.5.1C) , this strategy reduces the possibilities to the synthesis of a monomeric unit (4.5.1D) . The overall structure has S4 symmetry and is achiral even though assembled from chiral precursors. Both (+)-nonactic acid and (−)-nonactic acid (4.5.1D) are needed to construct the macrocycle and they are joined head-to-tail in an alternating (+)-(−)-(+)-(−) pattern. (see J. Am. Chem. Soc., 131, 17155 (2009) and references cited therein).

Is a part of the structure already solved? Critical study of the literature may often reveal that the same molecule or a closely related one has been solved. R.B. Woodward synthesized (4.5.2C) as a key intermediate in an elegant synthesis of Reserpine (4.5.2A) . The same intermediate compound (4.5.2C) became the key starting compound for Velluz et.al., in the synthesis of Deserpidine (4.5.2B) (Fig 4.5.2) .

Such strategies reduce the time taken for the synthesis of new drug candidates. These strategies are often used in natural product chemistry and drug chemistry. Once the preliminary scan is complete, the target molecule could be disconnected at Strategic Bonds.

4.6 Strategic Bonds, Retrons and Transforms

STRATEGIC BONDS are the bonds that are cleaved to arrive at suitable Starting Materials (SM) or SYNTHONS. For the purpose of bond disconnection, Corey has suggested that the structure could be classified according to the sub-structures generated by known chemical reactions. He called the sub-structures RETRONS and the chemical transformations that generate these Retrons were called TRANSFORMS. A short list of Transforms and Retrons are given below (TABLE 4.6.1). Note that when Transforms generate Retrons, the product may have new STEREOPOINTS (stereochemical details) generated that may need critical appraisal.

The structure of the target could be such that the Retron and the corresponding Transforms could be easily visualized and directly applied. In some cases, the Transforms or the Retrons may not be obvious. In several syntheses, transformations do not simplify the molecule, but they facilitate the process of synthesis. For example, a keto- group could be generated through modification of a -CH-N O 2 unit through a Nef reaction . This generates a new set of Retron / transforms pair. A few such transforms are listed below, along with the nomenclature suggested by Corey (Fig 4.6.2) .

A Rearrangement Reaction could be a powerful method for generating suitable new sub-structures. In the following example, a suitable Pinacol Retron, needed for the rearrangement is obtained through an acyloin transform (Fig 4.6.3) . Such rearrangement Retrons are often not obvious to inexperienced eyes.

Some transforms may be necessary to protect (acetals for ketones), modify (reduction of a ketone to alcohol to avoid an Aldol condensation during a Claisen condensation) or transpose a structural element such as a stereopoint (e.g. S N 2 inversion, epimerization etc.,) or shifting a functional group. Such transforms do not simplify the given structural unit. At times, activation at specific points on the structure may be introduced to bring about a C-C bond formation and later the extra group may be removed. For example, consider the following retrosynthesis in which an extra ester group has been introduced to facilitate a Dieckmann Retron. In complex targets, combinations of such strategies could prove to be a very productive strategy in planning retrosynthesis. Witness the chemical modification strategy shown below for an efficient stereospecific synthesis of a trisubstituted olefin (Fig 4.6.4)

Figure 4.6.4 Examples for FGA / FGR strategies for complex targets

Amongst the molecular architectures, the bridged-rings pose a complex challenge in Structure-Based disconnection procedures. Corey has suggested guidelines for efficient disconnections of strategic bonds.

A bond cleavage for retrosynthesis should lead to simplified structures, preferably bearing five- or six-membered rings. The medium and large rings are difficult to synthesize stereospecifically. Amongst the common rings, a six-membered ring is easily approached and manipulated to large and small rings. Simultaneous cleavage of two bonds, suggesting cycloaddition – retrons are often more efficient. Some cleavages of strategic bonds are shown in Figure 4.6.5, suggesting good and poor cleavage strategies based on this approach. However, these guidelines are not restrictive.

Identifying Retron – Transform sets in a given target molecule is therefore a critical component in retrosynthesis. Such an approach could often generate several synthetic routes. The merit of this approach is that starting materials do not prejudice this logic. Retrosyntheses thus developed could throws open several routes that need further critical scrutiny on the basis of known facts.

Identification of Retrons / Transforms sets provided the prerequisite for computer assisted programs designed for generating retrosynthetic routes. A list of Retrons and the corresponding transforms were interlinked and the data was stored in the computer. All known reactions were thus analyzed for their Retron / Transform characteristics and documented. The appropriate literature citations were also documented and linked. Based on these inputs, computer programs were designed to generate retrosynthetic routes for any given structure. Several such programs are now available in the market to help chemists generate synthetic strategies. Given any structure, these programs generate several routes. Once the scientist identifies the specific routes of interest for further analysis, the program generates detailed synthetic steps, reagents required and the appropriate citations. In spite of such powerful artificial intelligence, the intelligence and intuitive genius of a chemist is still capable of generating a new strategy, not yet programmed. Again, human intelligence is still a critical input for the analysis the routes generated using a computer. Based on the experience of the chemists’ team, their projected aim of the project and facilities available, the routes are further screened.

4.7 Elaboration of the concepts

Short lists of syntheses that exemplify retroanalysis strategies devised through powerful transforms are given below. Several syntheses from natural product chemistry are later discussed in this chapter, which further illustrates these points.

Retrosynthesis based on Diels-Alder Transform; (E.J. Corey et.al., J.A.C.S. (1972), 94, 2549). Fumagillol (4.7.1A) presents 4 stereocentres and sensitive functionalities.

Simplification of the functional groups first exposed a vic-diol. This site could come from an olefin D. Further retroanlysis led to a structurally simplified target sequence B to F. A cyclohexene ring system is suitable for a powerful DA Transform. This step generated two stereocentres in one reaction and also an olefin in the correct position for hydroxylation. The key intermediate C could also be generated through a functional group transform leading to G. This provided scope for a new set of starting materials using another DA Transform. The Retrosynthetic analysis and the actual synthesis are shown in Figure 4.7.1.

Synthetic protocol reported by Corey is outlined in Figure 4.7.2

For the synthesis of Estrone, an interesting DA Transform strategy was devised by Kametani et.al.. The retrosynthetic strategy is depicted in Fig 4.7.3. The required diene precursor was generated via cyclo-reversion reaction of a cyclobutene unit (T. Kametani et.al., Tetrahedron, (1981), 37, 3).

The crucial stereospecific trisubstituted olefins on Squalene (4.6.4B) were synthesized using a Claisen Retron 4.7.4A (Fig 4.7.4) . Note the double Claisen approach in this strategy.

The biogenetic-type cyclisation of olefins provides scope for application of Mechanistic Transform or transforms based on mechanistic considerations. A cleaver introduction of a chiral centre provided an efficient route for generating several enantiopure chiral centres in one step using this strategy (Fig 4.7.5) .

4.8 The problem of enantiomers

In these lengthy discussions above, we learnt about disconnection approaches. We said that stereocentres could introduce special challenges in planning efficient synthetic routes. Let us look at the molecule Biotin to understand disconnection strategies and problem of stereocentres.

Baker established the structure of Biotin in 1947 through an unambiguous synthesis of the molecule. A retroanalysis of the synthetic scheme is as shown below (Fig 4.8.1) .

The SM chosen and the synthetic approach clearly established the atom connectivities and the overall structure of the compound. However, the route made not attempt to synthesize one pure isomer because the actual stereochemistry was not established at that time. The route yielded all the eight stereoisomers (3 asymmetric centers). These isomers were carefully separated. In 1952 the biologically active isomer was identified as the all cis- enantiomer (+)-Biotoin. At this stage, several groups reported the stereospecific synthesis of the all cis- isomer exclusively (Fig 4.8.2) . The following retroanalysis depicts three such attempts. Note that these efforts were directed towards the synthesis of the racemate and not the pure (+)- isomer of Biotin.

These approaches solved the problem of diastereomeric purity. But they still left a mixture of two unresolved enentiomer viz., (±)-Biotin. To obtain a pure enantiomer in excellent yields, you have to resolve the racemic mixtures at appropriate stages. Alternately one could resort to asymmetric synthesis at all crucial stages. A still better approach would be to start from a chiral SM, which has most of the stereocentres in the correct fashion. This elegant approach is called the Chiron approach . When carefully executed, such procedures yield very pure enantiomer as the final product. Two such approaches for (+)-Biotin is shown below. In the first approach a chiral amino acid cysteine is chosen because it has one key asymmetric center, the sulphur moiety and a carboxylic acid in the correct positions (Fig 4.8.3) . In this example the choice of the SM is quite obvious. Note the cleaver introduction of the second nitrogen and the cyclisation step leading to the formation of the tetrahydrothiophene ring. Also note that the yield of Cram vs anti-Cram (chelation) products could be influenced by a choice of the reagent. These kinds of insights come only through a thorough knowledge of this particular reaction.

In the second example chosen here, the choice of the SM as the appropriate chiron is not obvious but hidden. Such an analysis demands a more critical insight into the concerned stereocentres.

An emerging concept in the Logic In Synthesis is deliberate planning of Green Synthetic Pathways. The logic of retroanalysis is same as discussed above. The only differentiating point is that the criteria for selection of synthetic route discussed earlier would now analyses the same synthetic tree through a Green Chemistry window to select only those routes that have maximum Green aspects. The green chemistry goal is enforced through inclusion the Twelve Principles of Green Chemistry. This could be done by embracing one or more of the following techniques - Use of Green Energy Sources like Microwave, Sonochemistry, Photochemistry etc., solvent free syntheses, using easily recoverable new solvent and eco-friendly solvents, reusable catalysts in syntheses and schemes that avoid protecting group chemistry. Most of the chemistry used in Green Chemistry is not really new to chemists. Chemistry is now revisited due to the environment consciousness that has now crept into industrial chemistry and society at large. Most of the chemistry is buried in two centuries of chemical literature. Several new discoveries in reagents have appeared in recent years. Now Chemists have to become more alert to this awakening to environmental damages caused by chemical activities on this globe.

The above discussions are meant only to illustrate the major steps involved in retrosynthetic analysis of a molecule. Thorough knowledge of synthetic tools, mechanisms and stereochemistry are essential prerequisites for a chemist to venture into the synthesis of complex molecules. Needless to add that all these efforts have to be suitably backed up by a team of chemists, having a rigorous training in laboratory techniques, a first-hand experience on several organic reactions / reagents and thorough knowledge of purification techniques, spectroscopic techniques and not the least, a good knowledge of search techniques to scan and retrieve requisite information from the vast chemical literature accumulated since the dawn of modern chemistry.

Retroanalysis of Some Interesting Molecules

Let us now dwell deep into a few select structures chosen from natural product chemistry and see how these structures have been tackled through different synthetic strategies. We would start with a simple molecule – Disparlure – with only two asymmetric centers. The course would end with a flovour of some Green Chemistry based syntheses to draw the attention of students to this newly emerging concepts and concerns.

• Search Citation
• Search Text
• Instructions for Articles
• Instructions for Perspectives
• Instructions for Discussion Addenda
• Instructions for Article Templates
• Instructions for Proposing Articles
• Article Template
• OS Perspective Template
• Discussion Addenda Template
• Ethical Guidelines
• Submit an Article
• Submit a Perspective
• Submit a Discussion Addendum
• Propose an Article
• Suggest a Subject for an Article
• About OrgSyn
• DEIR Statement
• History of OrgSyn
• Board of Editors
• Board of Directors
• Former Editors
• List of Volume Editors
• Most Accessed Articles
• Safety and Waste Disposal Information
• Safety Notes
• OS Philanthropy
• OS PUI Grants
• OS Workshop
• Website related questions
• Procedure related questions
• Use + button to add search terms
• Press operator buttons to rotate among Boolean operators AND, OR, and NOT.
• When more than one logical operator is used in a statement, NOT is evaluated first, then AND, and finally OR. Use ( or ) buttons to activate/deactivate parentheses to change the order of logical operators.
• To upload files please use structure editor "Open" menu button ;

Chemistry Matters—The Art of Organic Synthesis

9 • Chemistry Matters

If you think some of the synthesis problems at the end of this chapter are difficult, try devising a synthesis of vitamin B 12 starting only from simple substances you can buy in a chemical catalog. This extraordinary achievement was reported in 1973 as the culmination of a collaborative effort headed by Robert B. Woodward of Harvard University and Albert Eschenmoser of the Swiss Federal Institute of Technology in Zürich. More than 100 graduate students and postdoctoral associates contributed to the work, which took more than a decade to complete.

Why put such extraordinary effort into the laboratory synthesis of a molecule so easily obtained from natural sources? There are many reasons. On a basic human level, a chemist might be motivated primarily by the challenge, much as a climber might be challenged by the ascent of a difficult peak. Beyond the pure challenge, the completion of a difficult synthesis is also valuable in that it establishes new standards and raises the field to a new level. If vitamin B 12 can be made, then why can’t any molecule found in nature be made? Indeed, the decades that have passed since the work of Woodward and Eschenmoser have seen the laboratory synthesis of many enormously complex and valuable substances. Sometimes these substances—for instance, the anticancer compound paclitaxel, trade named Taxol—are not easily available in nature, so laboratory synthesis is the only method for obtaining larger quantities.

But perhaps the most important reason for undertaking a complex synthesis is that, in so doing, new reactions and new chemistry are discovered. It invariably happens in a complex synthesis that a point is reached at which the planned route fails. At such a time, the only alternatives are either to quit or to devise a way around the difficulty. New reactions and new principles come from such situations, and it is in this way that the science of organic chemistry grows richer. In the synthesis of vitamin B 12 , for example, unexpected findings emerged that led to the understanding of an entire new class of reactions—the pericyclic reactions that are the subject of Chapter 30 in this book. From synthesizing vitamin B 12 to understanding pericyclic reactions—no one could have possibly predicted such a link at the beginning of the synthesis, but that is the way of science.

As an Amazon Associate we earn from qualifying purchases.

This book may not be used in the training of large language models or otherwise be ingested into large language models or generative AI offerings without OpenStax's permission.

Want to cite, share, or modify this book? This book uses the Creative Commons Attribution-NonCommercial-ShareAlike License and you must attribute OpenStax.

Access for free at https://openstax.org/books/organic-chemistry/pages/1-why-this-chapter

• Authors: John McMurry, Professor Emeritus
• Publisher/website: OpenStax
• Book title: Organic Chemistry
• Publication date: Sep 20, 2023
• Location: Houston, Texas
• Book URL: https://openstax.org/books/organic-chemistry/pages/1-why-this-chapter
• Section URL: https://openstax.org/books/organic-chemistry/pages/9-chemistry-matters-the-art-of-organic-synthesis

© Jan 9, 2024 OpenStax. Textbook content produced by OpenStax is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike License . The OpenStax name, OpenStax logo, OpenStax book covers, OpenStax CNX name, and OpenStax CNX logo are not subject to the Creative Commons license and may not be reproduced without the prior and express written consent of Rice University.

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

• View all journals
• Explore content
• About the journal
• Publish with us
• Sign up for alerts
• Perspective
• Published: 12 June 2019

The digitization of organic synthesis

• Ian W. Davies 1  

Nature volume  570 ,  pages 175–181 ( 2019 ) Cite this article

27k Accesses

69 Citations

58 Altmetric

Metrics details

• Cheminformatics
• Process chemistry
• Chemical synthesis

Organic chemistry has largely been conducted in an ad hoc manner by academic laboratories that are funded by grants directed towards the investigation of specific goals or hypotheses. Although modern synthetic methods can provide access to molecules of considerable complexity, predicting the outcome of a single chemical reaction remains a major challenge. Improvements in the prediction of ‘above-the-arrow’ reaction conditions are needed to enable intelligent decision making to select an optimal synthetic sequence that is guided by metrics including efficiency, quality and yield. Methods for the communication and the sharing of data will need to evolve from traditional tools to machine-readable formats and open collaborative frameworks. This will accelerate innovation and require the creation of a chemistry commons with standardized data handling, curation and metrics.

You have full access to this article via your institution.

Similar content being viewed by others

Bridging the information gap in organic chemical reactions

Synthetic organic chemistry driven by artificial intelligence

Ratcheting synthesis

The preparation of oxalic acid and urea by Wöhler almost 200 years ago established the field that we call organic synthesis 1 . Human insight from reactivity explored in the interim can now lead to beautifully organized campaigns of complex natural products and bioactive molecules, which represent the pinnacle of synthetic design 2 . The idea of a synthesis machine that can build any molecule dates from the 1960s. However, although the first computer programs to design organic syntheses emerged around this time 3 , 4 , they failed to capture the imagination of chemists. Synthesis laboratories have remained sceptical of the ability of computer programs to learn the ‘art’ of organic chemistry, and have continued their tried and true approaches in their laboratories.

Now, the scepticism of synthetic chemists seems to be on the verge of changing. Using computer-aided synthesis planning (CASP), it is now possible to take the molecular structure of a desired product and output a detailed list of reaction schemes that connect the target molecule to known and often purchasable starting materials through a sequence of intermediates that are likely to be unknown 5 , 6 (Box 1 ). For example, the decision-tree-like search engine Chematica —which has a user-friendly graphical user interface and has been coded with human-curated rules over the past decade—has received laboratory validation of the predicted synthesis of medicinally relevant targets 7 . Approaches towards such programs usually reflect the priorities and prejudices of the programmers, and others have used different approaches—for example, using machine-learning algorithms or Monte Carlo Tree Search (as in AlphaGo 8 ) to guide the search, and a filter network to pre-select the most promising retrosynthetic steps that is trained on essentially all reactions ever published in organic chemistry 9 , 10 , 11 . In the future, it will be substantially faster for such programs to learn automatically from the primary data rather than rely on extracted rules and hand-designed heuristics, in analogy to the differences in strategy between Stockfish and AlphaZero in learning chess 12 .

The digitization of multistep organic synthesis is fast approaching, and the automation of the synthesis planning is just the first component that must be considered before automated reaction prediction can become a reality. The selection of reaction conditions is a key element of automated reaction prediction and is potentially a far more challenging task 13 (Fig. 1 ). This Perspective surveys the current prospects for the prediction of above-the-arrow conditions and addresses the challenges that are involved in integrating them into optimal methods of synthesis. For one, it has been stated that “syntheses are reported in prose” 14 . Not only are the reactions conditions often poorly communicated, but details are also omitted when explaining exactly how operations were carried out, meaning that many assumptions are made about the skills of the researcher repeating the synthesis. The prediction problem must then consider an even broader range of variables in order to master or fully execute a synthesis or optimization, depending on the context of academic research and medicinal or process chemistry.

To perform an organic chemical reaction in a laboratory, the conditions listed above the arrow are required to run the synthesis and isolate the desired product.

Box 1 Computer-aided synthesis planning

Computer-aided synthesis planning software was first described in the late 1960s 3 , 4 . Recently, machine-learning-based tools have been developed that provide information on route planning for a target molecule 5 , 6 . These algorithms are trained on the chemical literature, learning the ‘rules and reasoning’ of synthesis, and then predict a suitable synthetic route. They have been shown to be comparable to suggested routes from trained chemists towards medicinally relevant targets 7 .

A route is predicted from commercial materials to give the desired target molecule.

These critical advances in machine-aided synthesis are still limited in their application to more complex molecules such as natural products, as well as in dealing with the intricacies of medicinal and process chemistry. They rely on the datasets published in journal articles, which represent only a fraction of the raw data collected in a given research project or company portfolio. The continued advancement and proliferation of machine learning requires that methods of sharing and communicating information change and move to open collaborative frameworks with fully published machine readable datasets that are more transparent, contextualized and traceable.

Challenges in culture and data reporting

Proposing specific reactions to a given target on the basis of the literature and canonical rules may seem to be a mysterious and daunting task to most, but it is considered a routine activity for practitioners of organic synthesis who begin to grasp the principles as chemistry undergraduates 15 . Throughout a career these skills are improved, and the well-trained chemist often uses rules and patterns of chemical reactivity that they have developed by immersion in the field. With a new synthetic problem at hand, the chemist tries to compare it to a known one before making sense of it—a similar concept to that used by deep-learning algorithms. Historically, having spent a day reading the literature or conducting database searches, the chemist absorbs the precedents and sets off for the laboratory. Within a modern chemistry setting, predicting the starting point for experimentation—especially for complex molecular environments—is now challenging for even the best-educated of chemists. The yield and the selectivity (chemo-, regio-, diastereo- and enantioselectivity) of any transformation in the field of catalysis can be controlled by millions of permutations—including temperature, solvent, ligand and ancillary reagents—even before other metrics of quality are applied. Simply using a large number of experiments in (electronic) notebooks to select the above-the-arrow conditions has been unsuccessful so far, as the data are fractured and collected without diversity of starting materials—often because organizations have experience with molecules that were influenced by a target area in biology. Another obstacle related to human nature is that when reactions fail, the experimentalist is often not concerned with complete documentation and moves onto another task. In the area of medicinal chemistry, in which enormous numbers of experiments are performed, it has been stated that there are only two yields that matter: enough and not enough 16 . Overall, the current approaches used to record experiments fail to capture the ‘messiness’ of organic synthesis, as well as the continuous nature of the solutions in the real world.

To advance the field of machine learning in organic synthesis, enormous improvements will be required to enable the prediction of the discrete and continuous variables in the reaction conditions that appear above the arrow (Fig. 1 ). This will be possible only if it is accompanied by advances in the reporting of cases in which syntheses are captured in the form of digital code that can be published, versioned and transferred flexibly between platforms to enhance reproducibility. Despite the abundant incentives for academic and industrial scientists to share synthetic data via publication, the data published in most journal articles represents only a fraction of the raw data collected in a given research project. As a community we rely on outdated means that are mere facsimiles rather than machine-readable formats. A stumbling block is not only how to uniformly collect, clean and label data that are of use for training inside an organization or laboratory, but also how to align incentives to make data broadly available via new data intermediaries.

Further challenges for machine learning concern the identification and scoring of the criteria for the efficiency of the overall synthetic sequence, as there are currently no clear criteria on which this can be judged. It is already impossible for a human to assess all available options from either the recalling of synthetic methods or searching online. The formulation of such rules has primarily occurred in an academic setting around the definition of an ideal synthesis 17 , 18 , 19 . The ‘fit-for-purpose’ rule of academia or medicinal chemistry will certainly be unacceptable in the fine- and commodity-chemical sectors of the industry, in which efficiency, quality and safety are all a necessity. The reaction steps, time to a workable answer, speed and throughput, availability of diverse raw materials, process economics, sustainability and energy consumption all need to be included in assessing digitization of the multistep synthesis to define an answer that is beyond the output of a detailed list of potential reaction schemes.

Complexity in the execution of synthesis

The total synthesis of maoecrystal V (Fig. 2 ) is a good illustration of the level of above-the-arrow complexity in contemporary natural-product synthesis. In the preparation of this compound, which was completed by the Baran laboratory 20 , what is essentially an aldol reaction—taught in first-year organic chemistry classes—proved to be the most challenging step.

The natural product is prepared in a longest linear sequence of 11 steps. Step 7 is a reaction of enone 1 and formaldehyde to provide hydroxymethylketone 2 . In order to perform this reaction in a laboratory, at least 16 conditions—including workup procedures—are listed above the arrow.

The enolate-based installation of the hydroxymethyl group overcame the challenges of chemo- and regioselectivity. Over 1,000 experiments were carried out in order to optimize the reaction conditions, changing every conceivable variable possible; as a result, the optimized reaction has at least 16 conditions listed above the arrow. Conditions such as solvent and temperature changes and those used in workups are rarely considered in this context, but are essential for the successful repetition of the experiment. The desired product 2 was obtained with complete chemoselectivity, although the diastereoselectivity (2:1) and the yield (84%) remained intransigent to further improvement. Although far from optimal, the intermediate hydroxymethylketone 2 was processed onto maoecrystal V to provide sufficient material to answer the key biological questions presented by this molecule.

Different challenges prevail in the field of medicinal chemistry, in which molecules are designed to engage with increasingly more complex biological targets. Hundreds or thousands of molecules are required to advance from a hit compound to a drug candidate, and the synthetic route provides a platform from which to optimize for molecular function and explore biology. A consideration for any reaction used in medicinal chemistry is its level of tolerance to the polar functional groups and nitrogen heteroatoms that are typically found in biologically active molecules. As artificial intelligence and big data are increasingly used in medicinal chemistry for compound prediction and prioritization, it will become even more important to make the right compound the first time 21 . It is clear that even for well-precedented reactions and obvious retrosynthetic disconnections (that is, breaking a molecule up into simpler starting materials), there are fundamental practical limitations when considering the conditions needed to make sufficient material for biological testing 22 . Even within the context of the late-stage functionalization of a drug-like molecule, the individual conditions in that single step can still profoundly affect selectivity 23 .

As with natural-product synthesis, process chemistry has often been described as an ‘art’ 24 . Well-trained organic chemists read literature and generate the reaction sequence that in their best estimate meets their goals; however, these estimates are often biased by cultural- and company-based specific information on route selection, approaches to reject impurities, and the preparation of salts to improve the crystallinity, solubility and stability of intermediates or the active pharmaceutical compound. Process chemists have developed an intuition as to how well a reaction is likely to scale to obtain a high yield, high concentration, and low catalyst loading with good impurity rejection, and this informs the choice of a synthesis. This informal knowledge, acquired over many years by real-world reinforcement, is rarely captured in any form besides institutional knowledge.

Additionally, only a few well-conceived ideas can currently be pursued by process chemists in the laboratory. Commercial and regulatory pressures ensure that, among the range of potential routes identified early on, a single approach will be taken forward for validation and commercialization. These decisions are made largely with contradictory—or, at best, missing—data concerning the future potential efficiency of the route. This critical selection process is performed in the absence of quantitative efficiency data, and is often influenced by judgements on risk mitigation to product filing, or by broad assumptions around supply chain and tax and treasury. Although a considerable financial impact can be achieved by minimizing the costs of reagents and solvents and by optimizing the conditions for small improvements in yield or product quality, this impact cannot overcome the selection of a suboptimal route. It is highly desirable to understand all of the viable options before beginning full-scale development 25 .

Further predictions of conditions that are not historically included above the arrow must be used to narrow the range of options for further exploration. In process chemistry, the crystallinity and solubility, physical attributes of crystallization kinetics, particle-size reduction, flow ability, and solid-state stability are all key to understanding chemical intermediates and pharmaceutical properties. Thus, machine-learning algorithms should ideally be tailored to different criteria than those in other areas of synthesis. We will need to advance our ability to predict the organic-solvent solubility, the crystal phase and the morphology of compounds if we are to develop viable options without a priori knowledge.

Emerging examples of innovation using enhanced data

The goal of building a synthesis machine that can provide high-quality reagents for biology—beyond peptides and oligonucleotides—has been championed as a way of freeing up chemists for creative thinking by removing the bottleneck of synthesis 26 . However, a general commoditization of synthetic medicinal chemistry is not likely to emerge until we have made these orders-of-magnitude improvements in above-the-arrow prediction. Ultimately, machine learning will enable the field to predict individual conditions by moving along the spectrum of individual chemistry experiments, run one at a time, through large data assimilation and then back to individual conditions. A chemist can then, with a high degree of confidence, guarantee that sufficient product will be obtained in a single experiment to test the function of a molecule.

Scientists at Merck recognized this problem and systematically built tools, using high-throughput experimentation and analysis, to address the gaps in data 27 . Using the ubiquitous palladium-catalysed Suzuki–Miyaura cross-coupling reaction as a test case, they developed automation-friendly reactions that could operate at room temperature by using robotics employed in biotechnology coupled with emerging high-throughput analysis techniques. More than 1,500 chemistry experiments can be carried out in a day with this setup, using as little as 0.02 mg of starting material per reaction. This has since been expanded to allow for the in situ analysis of structure–activity relationships (nanoSAR) 28 . The authors note that, in the future, machine learning may aid the navigation of both reaction conditions and biological activity. Complementary approaches, such as inverse molecular design using machine learning, may also generate models for the rational design of prospective drugs 29 , 30 .

In order to reduce analysis time, ultra-high-throughput chemistry can be coupled to an advanced mass spectrometry method (such as matrix-assisted laser desorption ionization–time-of-flight spectrometry; MALDI–TOF) to enable the classification of thousands of experiments in minutes 31 . This classification approach may at first be slightly uncomfortable for synthetic chemists who hold stock in obtaining a hard yield, but it will surely become commonplace as more statistical methods and predictive models are deployed.

Machine learning has recently been used to predict the performance of a reaction on a given substrate in the widely used Buchwald–Hartwig C–N coupling reaction 32 . The Doyle laboratory used a robot-enabled simultaneous evaluation method with three 1,536-well plates that consisted of a full matrix of aryl halides, Buchwald ligands, bases and additives, giving a total of 4,608 reactions. The yields of these reactions were used as the model output and provided a clean, structured dataset containing substantially more reaction dimensions than have previously been examined with machine learning. Approximately 30% of the reactions failed to deliver any product, with the remainder spread relatively evenly over the range of non-zero yields. Using concepts popularized by the Sigman group 33 , scripts were built to compute and extract atomic, molecular and vibrational descriptors for the components of the cross-coupling. Using these descriptors as inputs and reaction yield as the output, a random forest algorithm was found to afford high predictive performance. This model was also successfully applied to sparse training sets and out-of-sample reaction outcome prediction, suggesting that a systematic reaction-profiling capability and machine learning will have general value for the survey and navigation of reaction space for other reaction types.

It has been suggested by Chuang and Keiser that this experimental design failed classical controls in machine learning, as it cannot distinguish chemically trained models from those trained on random features 34 . As they noted, flexible and powerful machine-learning models have become widespread, and their use can become problematic without some understanding of the underlying theoretical frameworks behind the models. The ability to distinguish peculiarities of the layout of an experiment from those that extract meaningful and actionable patterns also need to developed. Regardless, it is clear that the approach taken by Doyle—publishing a complete dataset and aligned code on GitHub—enables a clear demonstration of the scientific method of testing and generating hypotheses in independent laboratories.

The application of machine learning to the prediction of reactions has also been demonstrated for the conversion of alcohols to fluorides, the products of which are high-value targets in medicinal chemistry 35 (Fig. 2 ). In order to train a model for this reaction, descriptors for the substrates and reagents used in 640 screening reactions were tabulated. These included computed atomic and molecular properties as well as binary categorical identifiers (such as primary, secondary, cyclic). A random forest algorithm was used and was trained on 70% of the screening entries. The model was evaluated using a test set comprising the remaining 192 reactions and was validated on five structurally different substrates from outside the training set. The yields of these reactions were predicted with reasonable accuracy, which is more than sufficient to enable synthetic chemists to evaluate the feasibility of a reaction and to select initial reaction conditions. In comparison to previous studies, this training set was 80% smaller, encompassed much broader substrate diversity and incorporated multiple mechanisms. The expansion of the training set for this deoxyfluorination reaction to include additional variables (that is, stoichiometry, concentration, solvent and temperature) could lead to more accurate and comprehensive coverage of the complex reaction space.

Flow chemistry presents another opportunity for accelerated reaction development 36 . A recent publication by a Pfizer team 37 demonstrated high-throughput reaction screening of the Suzuki–Miyaura coupling with multiple discrete (catalyst, ligand and base) and continuous (temperature, residence time and pressure) variables (5,760 reactions in total), overcoming a common problem in which limited amounts of material do not allow for the application of flow reaction screening in medicinal chemistry (Fig. 3a, b ). Quinolines ( 3a – g ) and indazole acids ( 4a – d ) were used to validate the platform. In an important demonstration of the capability of the platform for the preparation of useful quantities of material, the team programmed the injection of 100 consecutive segments based on optimal conditions from screening, enabling the preparation of approximately 100 mg of a target molecule per hour.

Six hundred and forty screening reactions were performed to train a machine-learning model (yields presented as a heat map). This was used for the successful prediction of the yield and conditions for structurally different substrates that do not appear in the training set. This figure was adapted with permission from ref. 35 , copyright 2018 American Chemical Society.

The Jamison and Jensen groups have described an automated flow-based platform 38 to optimize above-the-arrow conditions to improve the yield, selectivity and reaction scope of a diverse range of reactions; this is typically a tedious and labour-intensive task in the laboratory. By using feedback from online analytics, the system converges on optimal conditions that can then be repeated or transferred with high fidelity as needed. These automated systems in academic laboratories may also play a part in the rapid collection of large, standardized datasets 39 .

Chemical synthesis may no longer be solely a human activity. In a recent study, the Cronin laboratory demonstrated that a robotic reaction-handling system controlled by a machine-learning algorithm might be able to explore organic reactions an order of magnitude faster than a manual process 40 . The robotic approach enabled the capture of information on failed or non-reactive experiments in a structured fashion, making it useful for reaction mapping. The powerful machine-learning algorithm was able to predict the reactivity of 1,000 reaction combinations from the above Pfizer dataset (Fig. 4a ), with greater than 80% accuracy, after considering the outcomes of around 10% of the dataset.

In this machine-learning analysis of the Pfizer work, one-hot encoding of the reaction conditions—in which the variables were assigned binary representations—and the clean standardized yield data were used to explore the prediction of yields by a neural network (catalyst loading and temperature were not included). In this approach, a random selection of 10% ( n  = 576) of the Suzuki–Miyaura reactions is used to train the neural net, and the remaining reactions are then scored by the model (Fig. 4b ). The candidates with the highest predicted yield are then added to the performed reactions, and the performance of the neural network is evaluated by calculating the mean of the true yield and the standard deviation of the yield. The neural network is then retrained, and the whole cycle is repeated until the entire space is explored in panels of 100 to demonstrate the alignment with the high-throughput experimentation as well as to evaluate the performance of the neural net. Such rapid evaluation is markedly enabled by the publication of reliable clean data.

A common theme in these three machine-learning examples is that predictions can be made with relatively small datasets: in some cases, with only 10% of the total number of reactions it is possible to predict the outcomes of the remaining 90%, without the need to physically conduct the experiments (Fig. 4 ). The high-fidelity data can originate from ultra-high-throughput screening, from flow chemistry or from an individual scientist, but the most important feature is the contextualized, internally consistent source that provides effective, secure and accurate data. This is important because it is currently not known how large these datasets need to be in order to predict across the molecules that represent drug-like space. Naturally, some reactivity trends may be reflective of how the individual experiments are conducted and not truly informative of a particular catalyst or ligand. A diagnostic approach using small libraries of curated drug-like molecules—known as ‘informer libraries’—has been presented as a way to better capture reaction scope and evolve synthetic models, but this should be viewed as an intermediary step as the field moves forward 22 .

a , A Suzuki–Miyaura reaction optimized in flow. A heat map of yields of the 5,760 reactions run is shown ( 3a – d with 4a – c and the reaction of 3e – g with 4d ), evaluated across a matrix of 11 ligands (plus one blank) × 7 bases (plus one blank) × 4 solvents (ref. 37 ). b , These data were used for one-hot encoding of reactants 3 , reactants 4 , ligands, bases and solvents as a test set for prediction of yield from the test set (30% of the reactions). Predictions for the full dataset are also shown. Panel a is adapted from ref. 37 , reprinted with permission from AAAS; panel b is adapted from ref. 40 .

There have also been important advances in predictive catalysis 41 , 42 . This is an exciting, emerging field that uses parameterization and analysis of catalysts to enable the forecast of an attainable improvement—for example, the enantioselectivity of a transformation or improved turnover in a biocatalytic reaction 43 , 44 , 45 —to provide confidence for route selection. For example, in the synthesis of letermovir 46 , a series of new catalysts was identified that provided the desired product in improved enantioselectivity and facilitated faster route optimization. The models are currently limited in scope, requiring a focused solvent screen on the best-performing catalysts, and process optimization had already taken place for the desired starting material. However, these models will greatly improve with the availability of enhanced datasets, which encompass a full range of activity from diverse sources 47 .

Extending these early successes to the prediction of the impurity profile of a reaction becomes especially difficult for catalysis, because many on-cycle and off-cycle events can markedly alter the optimum yield and because impurities do not always track with conversion. The current machine-learning systems do not yet take the mechanism of byproduct formation into account. However, process chemists will need information in order to predict and understand both the fate of impurities formed during each step in the process and where impurities are removed in the overall sequence; this is necessary not only to improve performance but also, and often more importantly, to meet regulatory requirements. Almost all of this information currently resides with corporations and is elusive internally and hidden externally. The messiness of data in our broad field of organic synthesis remains a challenge, and we should seek more engagement and demand more focused attention than we have in the past 50 years 48 .

Accelerating future innovation

There is a recent trend for organic chemists to publish ever larger numbers of examples in methodology papers. However, these reports remain focused on the knowledge and the dataset published in a journal article, which represents only a small portion of the raw data collected. These data have not yet been collected in a standardized manner, and highly complex substrates are often not included. In more general terms, in the 200-year history of organic synthesis, we have not yet developed methods to collect, clean and label data in a way that makes it useful for training in the context of new reaction optimization, especially in the areas of catalysis design and development. Existing datasets in the public or private domain have simply not been built with this in mind.

We have seen that large datasets or even ultra-high-throughput experimentation are not a prerequisite to machine learning. Biopharma deals with hundreds of millions of documents—including laboratory data and clinical trial reports, publications and patent filings, as well as billions of database records. Companies and not-for-profit alliances are working to provide solutions to data management. Despite the quantity of data, chemical structure information is essentially captured as an image in a book—it is essentially unusable, whereas above-the-arrow and other data are currently considered out of scope and the vast amounts of historic data in paper and electronic notebooks remains orphaned. Consequently, to avoid repeating current synthetic methods in the field we need to embrace modern approaches and pay attention to future needs. This will avoid simply restating the master data problem. Metrics for similarity calculations 49 use the fingerprints of molecules to compare how similar they are to each other and will ensure that we avoid bias introduced by human-curated examples for machine learning. We need our data to emerge beyond the positive results and the publication- or career-driven biases. A published data point should be one click away from raw experimental data, all the way from the weighing of materials to analytical data, enabled by the Internet of Things 50 .

Before we do that, we need to provide a framework in which to enable the collection and publication of new data as it is generated, much like the Bermuda Accord 51 . This established that all the DNA sequence information from large-scale human genomic projects should be freely available and in the public domain. With increasing exploration of new research areas that cross disciplines—for example, chemical biology or proteomics—it is becoming common for very different traditions towards data sharing to coexist in the same laboratory 52 . In the field of organic synthesis the intensity of the work, the amount of capital allocation required and the degree of specialization in data rather than ‘art’ will lead to the creation of a new kind of chemist—one whose principal objective is the generation of high-quality datasets. These datasets will go on to be the foundation for a new partnership of hypothesis-driven and hypothesis-free discovery based on big data in chemistry 53 . This distinction exists today in biology as the number of data-generating projects advances, in which medical breakthroughs such as CRISPR often emerge from unpredictable origins. The field has adapted, and enables academic data-generating researchers to continue obtaining grant funding for their work as well as advancing their careers through publication and peer recognition. Governmental agencies and large independent global charities can clearly influence the funding of the new data-generation projects, science policy, intellectual property and regulation.

Synthetic chemistry has emerged relatively unscathed from the narrative of poor reproducibility in science and has not yet faced a crisis of confidence 54 . There have been important calls regarding reproducibility 55 and the discussion will remain contemporary as it essential that the quality, reproducibility and traceability of the raw data and models. As in several of the machine-learning examples discussed above 32 , 40 , the availability of reliable data and the code enables others to verify and retest alternative hypotheses. This helps to demonstrate the effect of data that is findable, accessible, interoperable and reusuable (FAIR) 56 . For example, the Cronin group was able to rapidly model the data for the Suzuki–Miyaura reaction presented from the Pfizer flow platform 40 . As we report more complete datasets that include reactions that fail to give products in expected yield or quality, we need to be cautious. A failure may not represent a true reflection of the reactivity profile of a current method, and we need to ensure that it does not limit exploration or utilization of a newly developed reaction. In the future, machine learning will therefore need to become a partner in order to elucidate reaction concepts, elusive high-value transformations and problems of which chemists are not currently aware (unknown unknowns), as well as to rapidly identify unanticipated observations or spare events.

It is exciting to consider the potential societal impact of innovations similar to AlphaGo Zero in the chemical space. Commercial software packages are emerging and, although it is clear that these approaches will advance in sophistication, it is not necessary for the end user to understand the underlying complexity as long as the answers satisfy their needs. Unlike in closed systems such as chess or Go, there are no clearly defined rules for winning, and explainable artificial intelligence will be an ongoing issue 57 , 58 . It remains to be seen whether the machines can become experts or merely expert tools.

Future advances in the digitization of chemistry will not come at an equal pace, and some areas of organic synthesis will be affected much sooner than others. Computing power is no longer a limitation, and there are much more sophisticated algorithms that can handle fuzzy datasets developing in fields that have more direct monetization. Although the technology is not yet reliable enough, it is clear that the field of synthesis and optimization in applications such as medicinal and process chemistry will become a more evidence-led practice. Some organic chemists will ignore the signals of this transformation, some will improve and make incremental progress, and some will be the innovators, embracing these tools to augment their scientific intuition and creativity.

Wöhler, F. Ueber künstliche bildung des harnstoffs. Ann. Phys . 88 , 253–256 (1828).

Article   Google Scholar  

Whitesides, G. M. Complex organic synthesis: structure, properties, and/or function? Isr. J. Chem . 58 , 142 (2018).

Article   CAS   Google Scholar  

Corey, E. J. & Wipke, W. T. Computer-assisted design of complex organic syntheses. Science 166 , 178–192 (1969).

Article   CAS   ADS   Google Scholar  

Corey, E. J., Wipke, W. T., Cramer, R. D. III & Howe, W. J. Computer-assisted synthetic analysis. Facile man–machine communication of chemical structure by interactive computer graphics J. Am. Chem. Soc . 94 , 421–430 (1972).

Szymkuć, S. et al. Computer-assisted synthetic planning: the end of the beginning. Angew. Chem. Int. Ed . 55 , 5904–5937 (2016).

Coley, C. W., Green, W. H. & Jensen, K. F. Machine learning in computer-aided synthesis planning. Acc. Chem. Res . 51 , 1281–1289 (2018).

Klucznik, T. et al. Efficient syntheses of diverse, medicinally relevant targets planned by computer and executed in the laboratory. Chem 4 , 522–532 (2018).

Silver, D. et al. Mastering the game of Go without human knowledge. Nature 550 , 354–359 (2017).

Schwaller, P., Gaudin, T., Lanyi, D., Bekas, C. & Laino, T. “Found in translation”: predicting outcomes of complex organic chemistry reactions using neural sequence-to-sequence models. Chem. Sci . 9 , 6091–6098 (2018).

Segler, M. H. S. & Waller, M. P. Modelling chemical reasoning to predict and invent reactions. Chem. Eur. J . 23 , 6118–6128 (2017).

Segler, M. H. S., Preuss, M. & Waller, M. P. Planning chemical syntheses with deep neural networks and symbolic AI. Nature 555 , 604–610 (2018).

Kasparov, G. Chess, a Drosophila of reasoning. Science 362 , 1087 (2018).

Cernak, T. A machine with chemical intuition. Chem 4 , 401–403 (2018).

Steiner, S. et al. Organic synthesis in a modular robotic system driven by a chemical programming language. Science 363 , eaav2211 (2019).

Garg, N. K. Empowering students to innovate: engagement in organic chemistry teaching. Angew. Chem. Int. Ed . 57 , 15612–15613 (2018).

Engkvist, O. et al. Computational prediction of chemical reactions: current status and outlook. Drug Discov. Today 23 , 1203–1218 (2018).

Gaich, T. & Baran, P. S. Aiming for the ideal synthesis. J. Org. Chem . 75 , 4657–4673 (2010).

Trost, B. M. The atom economy—a search for synthetic efficiency. Science 254 , 1471–1477 (1991).

Burns, N. Z., Baran, P. S. & Hoffmann, R. W. Redox economy in organic synthesis. Angew. Chem. Int. Ed . 48 , 2854–2867 (2009).

Cernijenko, A., Risgaard, R. & Baran, P. S. 11-step total synthesis of (−)-maoecrystal V. J. Am. Chem. Soc . 138 , 9425–9428 (2016).

Griffen, E. J., Dossetter, A. G., Leach, A. G. & Montague, S. Can we accelerate medicinal chemistry by augmenting the chemist with Big Data and artificial intelligence? Drug Discov. Today 23 , 1373–1384 (2018).

Kutchukian, P. S. et al. Chemistry informer libraries: a chemoinformatics enabled approach to evaluate and advance synthetic methods. Chem. Sci . 7 , 2604–2613 (2016).

Yao, H. et al. Enabling efficient late-stage functionalization of drug-like molecules with LC-MS and reaction-driven data processing. Eur. J. Org. Chem . 2017 , 7122–7126 (2017).

Yasuda, N. (ed.) The Art of Process Chemistry (Wiley-VCH, 2010).

Li, J., Albrecht, J., Borovika, A. & Eastgate, M. D. Evolving green chemistry metrics into predictive tools for decision making and benchmarking analytics. ACS Sustainable Chem. Eng . 6 , 1121–1132 (2018).

Trobe, M. & Burke, M. D. The molecular industrial revolution: automated synthesis of small molecules. Angew. Chem. Int. Ed . 57 , 4192–4214 (2018).

Buitrago Santanilla, A. et al. Nanomole-scale high-throughput chemistry for the synthesis of complex molecules. Science 347 , 49–53 (2015).

Gesmundo, N. et al. Nanoscale synthesis and affinity ranking. Nature 557 , 228–232 (2018).

Sanchez-Lengeling, B. & Aspuru-Guzik, A. Inverse molecular design using machine learning: generative models for matter engineering. Science 361 , 360–365 (2018).

Schneider, G. Automating drug discovery. Nat. Rev. Drug Discov . 17 , 97–113 (2018).

Lin, S. et al. Mapping the dark space of chemical reactions with extended nanomole synthesis and MALDI-TOF MS. Science 361 , eaar6236 (2018).

Ahneman, D. T., Estrada, J. G., Lin, S., Dreher, S. D. & Doyle, A. G. Predicting reaction performance in C–N cross-coupling using machine learning. Science 360 , 186–190 (2018). This article demonstrates machine learning in prediction of the performance of a catalytic reaction using data obtained via high-throughput experimentation .

Zhao, S. et al. Enantiodivergent Pd-catalyzed C–C bond formation enabled through ligand parameterization. Science 362 , 670–674 (2018).

Chuang, K. V. & Keiser, M. J. Comment on “Predicting reaction performance in C–N cross-coupling using machine learning”. Science 362 , eaat8603 (2018). This article illustrates the need to incorporate random-control procedures when applying machine learning to new scientific domains and the importance of experimental design .

Article   ADS   Google Scholar  

Nielsen, M. K., Ahneman, D. T., Riera, O. & Doyle, A. G. Deoxyfluorination with sulfonyl fluorides: navigating reaction space with machine learning. J. Am. Chem. Soc . 140 , 5004–5008 (2018). This paper demonstrates the use of machine learning on a relatively small dataset obtained by traditional laboratory experimentation .

Reizman, B. J. & Jensen, K. F. Feedback in flow for accelerated reaction development. Acc. Chem. Res . 49 , 1786–1796 (2016).

Perera, D. et al. A platform for automated nanomole-scale reaction screening and micromole-scale synthesis in flow. Science 359 , 429–434 (2018). This article illustrates that a flow apparatus can accelerate reaction optimization earlier in the drug-discovery process and also provides reliable data that enables other laboratories to build machine-learning algorithms .

Bedard, A.-C. et al. Reconfigurable system for automated optimization of diverse chemical reactions. Science 361 , 1220–1225 (2018).

Caramelli, D. et al. Networking chemical robots for reaction multitasking. Nat. Commun . 9 , 3406 (2018).

Granda, J. M., Donina, L., Dragone, V., Long, D.-L. & Cronin, L. Controlling an organic synthesis robot with machine learning to search for new reactivity. Nature 559 , 377–381 (2018). This article predicts the reactivity of about 1 , 000 reaction combinations with accuracy greater than 80 per cent after considering the outcomes of slightly over 10 per cent of the dataset and , notably , the approach was also used to calculate the reactivity of published datasets .

Harper, K. C. & Sigman, M. S. Predicting and optimizing asymmetric catalyst performance using the principles of experimental design and steric parameters. Proc. Natl Acad. Sci. USA 108 , 2179–2183 (2011).

Zahrt, A. F. et al. Prediction of higher-selectivity catalysts by computer-driven workflow and machine learning. Science 363 , eaau5631 (2019).

Matsuda, T. (ed.) Future Directions in Biocatalysis 2nd edn (Elsevier, 2017).

Kan, S. B. J., Russell, D., Lewis, R. D., Chen, K. & Arnold, F. H. Directed evolution of cytochrome c for carbon–silicon bond formation: bringing silicon to life. Science 354 , 1048–1051 (2016).

Arnold, F. H. Innovation by evolution: bringing new chemistry to life – Nobel lecture. Nobel Media AB 2019 https://www.nobelprize.org/prizes/chemistry/2018/arnold/lecture/ (2019).

Metsänen, T. T. et al. Combining traditional 2D and modern physical organic-derived descriptors to predict enhanced enantioselectivity for the key aza-Michael conjugate addition in the synthesis of Prevymis™ (letermovir). Chem. Sci . 9 , 6922–6927 (2018).

Gedeck, P., Skolnik, S. & Rodde, S. Developing collaborative QSAR models without sharing structures. J. Chem. Inf. Model . 57 , 1847–1858 (2017).

Donoho, D. 50 years of data science. J. Comput. Graph. Stat . 26 , 745–766 (2017).

Article   MathSciNet   Google Scholar  

Bajusz, D., Racz, A. & Heberger, K. Why is Tanimoto index an appropriate choice for fingerprint-based similarity calculations? J. Cheminf . 7 , 20 (2015).

Martinot, T. Could Internet-of-Things be the next step in the evolution of chemistry. TetraScience Blog https://blog.tetrascience.com/blog/could-internet-of-things-be-the-next-step-in-the-evolution-of-chemistry/ (2016).

Contreras, J. L. Bermuda’s legacy: policy, patents, and the design of the genome commons. Minn. J. Law Sci. Technol . 12 , 61–125 (2011).

Google Scholar  

Amann, R. I. et al. Toward unrestricted use of public genomic data. Science 363 , 350–352 (2019).

Lander, E. S. The heroes of CRISPR. Cell 164 , 18–28 (2016).

Baker, M. Is there a reproducibility crisis? Nature 533 , 452–454 (2016).

Bergman, R. G. & Danheiser, R. L. Reproducibility in chemical research. Angew. Chem. Int. Ed . 55 , 12548–12549 (2016).

Brock, J. “A love letter to your future self”: what scientists need to know about FAIR data. Nature Index https://www.natureindex.com/news-blog/what-scientists-need-to-know-about-fair-data (2019).

Preece, A., Harborne, D., Braines, D., Tomsett, R. & Chakraborty, S. Stakeholders in explainable AI. Preprint at https://arxiv.org/abs/1810.00184 (2018).

Caliskan, A., Bryson, J. J. & Narayanan, A. Semantics derived automatically from language corpora contain human-like biases. Science 356 , 183–186 (2017).

Download references

Reviewer information

Nature thanks Ian Churcher, Jacob Janey and the other anonymous reviewer(s) for their contribution to the peer review of this work.

Author information

Authors and affiliations.

Princeton Catalysis Initiative, Princeton University, Princeton, NJ, USA

Ian W. Davies

You can also search for this author in PubMed   Google Scholar

Corresponding author

Correspondence to Ian W. Davies .

Ethics declarations

Competing interests.

The author declares no competing interests.

Additional information

Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Reprints and permissions

About this article

Cite this article.

Davies, I.W. The digitization of organic synthesis. Nature 570 , 175–181 (2019). https://doi.org/10.1038/s41586-019-1288-y

Download citation

Received : 20 August 2018

Accepted : 26 April 2019

Published : 12 June 2019

Issue Date : 13 June 2019

DOI : https://doi.org/10.1038/s41586-019-1288-y

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

This article is cited by

Prediction of chemical reaction yields with large-scale multi-view pre-training.

Journal of Cheminformatics (2024)

An integrated self-optimizing programmable chemical synthesis and reaction engine

• Artem I. Leonov
• Alexander J. S. Hammer
• Leroy Cronin

Nature Communications (2024)

An intriguing vision for transatlantic collaborative health data use and artificial intelligence development

• Daniel C. Baumgart

npj Digital Medicine (2024)

A deep learning framework for accurate reaction prediction and its application on high-throughput experimentation data

• Hongming Chen

Journal of Cheminformatics (2023)

Automate and digitize

Nature Synthesis (2023)

By submitting a comment you agree to abide by our Terms and Community Guidelines . If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.

Quick links

• Explore articles by subject
• Guide to authors
• Editorial policies

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.