Integrating Synthetic Accessibility with AI-based Generative Drug Design

Generative models are frequently used for de novo design in drug discovery projects to propose new molecules. However, the question of whether or not the generated molecules can be synthesized is not systematically taken into account during generation, even though being able to synthesize the generated molecules is a fundamental requirement for such methods to be useful in practice. Methods have been developed to estimate molecule synthesizability, but, so far, there is no consensus on whether or not a molecule is synthesizable. In this paper we introduce the Retro-Score (RScore), which computes a synthetic feasibility score of molecules by performing a full retrosynthetic analysis through our data-driven synthetic planning software Spaya, and its dedicated API: Spaya-API (https://spaya.ai). After a comparison of RScore with other synthetic scores from the literature, we describe a pipeline to generate molecules that validate a list of targets while still being easy to synthesize. We further this idea by performing experiments comparing molecular generator outputs across a range of constraints and conditions. We show that the RScore can be learned by a Neural Network, which leads to a new score: RSPred. We demonstrate that using the RScore or RSPred as a constraint during molecular generation enables to obtain more synthesizable solutions, with higher diversity. The open-source Python code containing all the scores and the experiments can be found on https://github.com/iktos/generation-under-synthetic- constraint.

Integrating Synthetic Accessibility with AI-based Generative Drug Design

Generative models are frequently used for de novo design in drug discovery projects to propose new molecules. However, the question of whether or not the generated molecules can be synthesized is not systematically taken into account during generation, even though being able to synthesize the generated molecules is a fundamental requirement for such methods to be useful in practice. Methods have been developed to estimate molecule synthesizability, but, so far, there is no consensus on whether or not a molecule is synthesizable. In this paper we introduce the Retro-Score (RScore), which computes a synthetic feasibility score of molecules by performing a full retrosynthetic analysis through our data-driven synthetic planning software Spaya, and its dedicated API: Spaya-API (https://spaya.ai). After a comparison of RScore with other synthetic scores from the literature, we describe a pipeline to generate molecules that validate a list of targets while still being easy to synthesize. We further this idea by performing experiments comparing molecular generator outputs across a range of constraints and conditions. We show that the RScore can be learned by a Neural Network, which leads to a new score: RSPred. We demonstrate that using the RScore or RSPred as a constraint during molecular generation enables to obtain more synthesizable solutions, with higher diversity. The open-source Python code containing all the scores and the experiments can be found on https://github.com/iktos/generation-under-synthetic- constraint.

The Total Synthesis of Aigialomycin D and Analogues

<p>In 2002 a new family of 14-membered resorcylic macrolides, the aigialomycins, were isolated from the mangrove fungus Aigialus parvus BCC 5311. Subsequent biological testing of these new natural products found aigialomycin D (Am D) to be the most biologically active member of the family, exhibiting moderate activity against malaria (Plasmodium falciparum K1, IC50 19.7 Mu) and modest cytotoxicity towards certain cancer cells (KB cells: IC50 9.0 Mu and BC-1 cells: 53.8 Mu). More recently, Am D has been shown to inhibit the kinases CDK1/5 and GSK at low Mu concentrations. At the onset of this research project, with only one total synthesis of Am D reported in the literature, there remained a need for an efficient synthesis of Am D that would be amenable to the synthesis of a range of analogues. This thesis reports two synthetic approaches to Am D that differ primarily in the chemistry utilised to install the (E)-olefins at C1'-C2' and C7'-C8': a Horner-Wadsworth-Emmons (HWE) strategy and a Ramberg-Backlund (RB) strategy. The Ramberg-Backlund strategy ultimately proved to be successful, providing Am D in 16 steps with 9% overall yield. A retrosynthetic analysis of Am D disconnects the molecule into three major fragments: an aromatic fragment, a C2'-C7' carbohydrate-derived fragment and a C8'-C11' alcohol fragment. The synthesis of the three fragments for each strategy is described and the attempts made to couple the fragments together, first with HWE methodology and then successfully with ring-closing metathesis (RCM) and RB reactions, are discussed. The synthesis of several Am D analogues and their preliminary biological testing is also described.</p>

The Total Synthesis of Aigialomycin D and Analogues

<p>In 2002 a new family of 14-membered resorcylic macrolides, the aigialomycins, were isolated from the mangrove fungus Aigialus parvus BCC 5311. Subsequent biological testing of these new natural products found aigialomycin D (Am D) to be the most biologically active member of the family, exhibiting moderate activity against malaria (Plasmodium falciparum K1, IC50 19.7 Mu) and modest cytotoxicity towards certain cancer cells (KB cells: IC50 9.0 Mu and BC-1 cells: 53.8 Mu). More recently, Am D has been shown to inhibit the kinases CDK1/5 and GSK at low Mu concentrations. At the onset of this research project, with only one total synthesis of Am D reported in the literature, there remained a need for an efficient synthesis of Am D that would be amenable to the synthesis of a range of analogues. This thesis reports two synthetic approaches to Am D that differ primarily in the chemistry utilised to install the (E)-olefins at C1'-C2' and C7'-C8': a Horner-Wadsworth-Emmons (HWE) strategy and a Ramberg-Backlund (RB) strategy. The Ramberg-Backlund strategy ultimately proved to be successful, providing Am D in 16 steps with 9% overall yield. A retrosynthetic analysis of Am D disconnects the molecule into three major fragments: an aromatic fragment, a C2'-C7' carbohydrate-derived fragment and a C8'-C11' alcohol fragment. The synthesis of the three fragments for each strategy is described and the attempts made to couple the fragments together, first with HWE methodology and then successfully with ring-closing metathesis (RCM) and RB reactions, are discussed. The synthesis of several Am D analogues and their preliminary biological testing is also described.</p>

Towards a Total Synthesis of Peloruside A and the Preparation of Selected Inositol Derivatives

<p>This thesis covers two broad areas of work under the general theme of the synthesis of bioactive and/or synthetically useful compounds based on natural products or deriving from the chiral pool. Chapters one, two and three focus on the marine secondary metabolite peloruside A (1), which has been shown to stabilise microtubules during mitosis and hence cause apoptosis (cell death) in a similar manner to the very successful anticancer drug Taxol. A synthetic program with the aim of devising a total synthesis was initiated at Victoria University of Wellington after peloruside A's discovery in 1999. Four synthetic disconnects were identified in the retrosynthetic analysis of peloruside A: to give the C-l to C-2 fragment; the C-3 to C-7 fragment; the C-8 to C-11 fragments; and the C-12 to C-24 fragment. The C-7 to C-8 bond was to be formed via an asymmetric aldol reaction to give the pyranose ring fragment (highlighted in blue). In this thesis, the synthesis of the C-3 to C-7 fragment is described. A1do1 reactions with the C-8 to C- 11 ketone have been investigated, and subsequent progress towards the assembly of the pyranose ring fragment is presented. Chapters four, five, six and seven describe the preparation of selected synthetically and biologically useful derivatives of the commercially available inositols, quebrachitol (L-chiro-inositol-2-methyl ether) and myo-inositol. The butane di-acetal (BDA) derivatives 293, 300, and 301 (as well as acetylated and methylated derivatives thereof) were prepared during work directed towards the synthesis of the inositol core of a phosphatidylinositol manno-oligosaccharide (PIM-6) isolated from Mycobacterium bovis and M. smegmatis. Quebrachitol derivatives 305, 306 and 307 were prepared and subsequently tested against myoinositol (the optimal competitor) in biological uptake assays of the microorganisms, Candida albicans and Leishmania donovani. For both microorganisms, the mono- and di-O-methylated L-chiro-inositol derivatives 307 and 305, as well as quebrachitol, gave significant inhibition results, with P values from P < 0.001 to P < 0.05 for paired-sample t-test analyses, i.e.99.9% to 95% confidence for significant inhibition, respectively. The benzoylated derivative 306 did not induce any inhibition of myo-inositol uptake. Myo-inositol is the most abundant of the inositols in nature and is readily available. However, as it is a meso compound, one of the key challenges in the use of myoinositol as a synthetic precursor is an efficient resolution method. The formation of myo-inositol camphanylidene acetal 269a is one successful solution, and work done in an attempt to better understand the selectivity of the reaction is reported here. Also, process development work was done to adapt the preparation so that it was suitable for scale-up, and a subsequent large scale synthesis of the acetal was undertaken. Previously unpublished X-ray crystal structures were obtained for 269a and, for two of the diastereomeric impurities of the reaction.</p>

Towards a Total Synthesis of Peloruside A and the Preparation of Selected Inositol Derivatives

<p>This thesis covers two broad areas of work under the general theme of the synthesis of bioactive and/or synthetically useful compounds based on natural products or deriving from the chiral pool. Chapters one, two and three focus on the marine secondary metabolite peloruside A (1), which has been shown to stabilise microtubules during mitosis and hence cause apoptosis (cell death) in a similar manner to the very successful anticancer drug Taxol. A synthetic program with the aim of devising a total synthesis was initiated at Victoria University of Wellington after peloruside A's discovery in 1999. Four synthetic disconnects were identified in the retrosynthetic analysis of peloruside A: to give the C-l to C-2 fragment; the C-3 to C-7 fragment; the C-8 to C-11 fragments; and the C-12 to C-24 fragment. The C-7 to C-8 bond was to be formed via an asymmetric aldol reaction to give the pyranose ring fragment (highlighted in blue). In this thesis, the synthesis of the C-3 to C-7 fragment is described. A1do1 reactions with the C-8 to C- 11 ketone have been investigated, and subsequent progress towards the assembly of the pyranose ring fragment is presented. Chapters four, five, six and seven describe the preparation of selected synthetically and biologically useful derivatives of the commercially available inositols, quebrachitol (L-chiro-inositol-2-methyl ether) and myo-inositol. The butane di-acetal (BDA) derivatives 293, 300, and 301 (as well as acetylated and methylated derivatives thereof) were prepared during work directed towards the synthesis of the inositol core of a phosphatidylinositol manno-oligosaccharide (PIM-6) isolated from Mycobacterium bovis and M. smegmatis. Quebrachitol derivatives 305, 306 and 307 were prepared and subsequently tested against myoinositol (the optimal competitor) in biological uptake assays of the microorganisms, Candida albicans and Leishmania donovani. For both microorganisms, the mono- and di-O-methylated L-chiro-inositol derivatives 307 and 305, as well as quebrachitol, gave significant inhibition results, with P values from P < 0.001 to P < 0.05 for paired-sample t-test analyses, i.e.99.9% to 95% confidence for significant inhibition, respectively. The benzoylated derivative 306 did not induce any inhibition of myo-inositol uptake. Myo-inositol is the most abundant of the inositols in nature and is readily available. However, as it is a meso compound, one of the key challenges in the use of myoinositol as a synthetic precursor is an efficient resolution method. The formation of myo-inositol camphanylidene acetal 269a is one successful solution, and work done in an attempt to better understand the selectivity of the reaction is reported here. Also, process development work was done to adapt the preparation so that it was suitable for scale-up, and a subsequent large scale synthesis of the acetal was undertaken. Previously unpublished X-ray crystal structures were obtained for 269a and, for two of the diastereomeric impurities of the reaction.</p>

Studies Towards the Synthesis of Peloruside A

<p>In the search for new treatments for cancer, advances in biology have provided targets for the destruction of cancer cells. One such structure the microtubule, a protein required for cell division, has been the target of many successful anticancer agents including the multi-million dollar earning Taxol [trademark] (paclitaxel) and the epothilones, currently in late-stage clinical trials. More recently it has been shown that peloruside A 1, a secondary metabolite isolated from the New Zealand marine sponge Mycale hentscheli, prevents cell division by stabilising microtubules, and thus offers promise as a novel anticancer agent. However, due to its limited natural abundance, significant quantities of peloruside A can only be obtained through chemical synthesis. A retrosynthetic analysis of peloruside A divided the molecule into four key fragments: a) the commercially available C-l to C-2 benzyloxy acetic acid fragment; b) the C-3 to C-7 fragment; c) the C-8 to C-11 fragment and d) the remaining C-12 to C-24 portion of the macrocycle and side chain. The C-3 to C-7 and C-8 to C-11 fragments combine to form a key intermediate pyranose ring. This thesis however, addresses the synthesis of two of these key fragments, namely the C-8 to C-11 and C-12 to C-24 fragments. An efficient synthesis of the C-8 to C-11 fragment of peloruside A, starting from commercially available pantolactone, has been developed. This synthesis proceeds in good overall yield, and has been successfully reproduced on the multigram scale. The significant portion of this thesis, however, is dedicated to the synthesis of the C-12 to C-24 fragment. After our initial strategy proved unviable, a short, facile method for the synthesis of the C-12 to C-24 fragment, involving the formation of a bis-silyl ether, was developed. The protocol for its desired coupling, via a boron_mediated, remote 1,5-anti-induction aldol reaction has also been established. These and subsequent studies provided valuable insight into the origin of 1,5-anti induction in boron-mediated aldol reactions.</p>

Studies Towards the Synthesis of Peloruside A

<p>In the search for new treatments for cancer, advances in biology have provided targets for the destruction of cancer cells. One such structure the microtubule, a protein required for cell division, has been the target of many successful anticancer agents including the multi-million dollar earning Taxol [trademark] (paclitaxel) and the epothilones, currently in late-stage clinical trials. More recently it has been shown that peloruside A 1, a secondary metabolite isolated from the New Zealand marine sponge Mycale hentscheli, prevents cell division by stabilising microtubules, and thus offers promise as a novel anticancer agent. However, due to its limited natural abundance, significant quantities of peloruside A can only be obtained through chemical synthesis. A retrosynthetic analysis of peloruside A divided the molecule into four key fragments: a) the commercially available C-l to C-2 benzyloxy acetic acid fragment; b) the C-3 to C-7 fragment; c) the C-8 to C-11 fragment and d) the remaining C-12 to C-24 portion of the macrocycle and side chain. The C-3 to C-7 and C-8 to C-11 fragments combine to form a key intermediate pyranose ring. This thesis however, addresses the synthesis of two of these key fragments, namely the C-8 to C-11 and C-12 to C-24 fragments. An efficient synthesis of the C-8 to C-11 fragment of peloruside A, starting from commercially available pantolactone, has been developed. This synthesis proceeds in good overall yield, and has been successfully reproduced on the multigram scale. The significant portion of this thesis, however, is dedicated to the synthesis of the C-12 to C-24 fragment. After our initial strategy proved unviable, a short, facile method for the synthesis of the C-12 to C-24 fragment, involving the formation of a bis-silyl ether, was developed. The protocol for its desired coupling, via a boron_mediated, remote 1,5-anti-induction aldol reaction has also been established. These and subsequent studies provided valuable insight into the origin of 1,5-anti induction in boron-mediated aldol reactions.</p>

MegaSyn: Integrating Generative Molecule Design, Automated Analog Designer and Synthetic Viability Prediction

Drug discovery is a multi-stage process, often beginning with the identification of active molecules from a high-throughput screen or machine learning model. Once structure activity relationship trends become well established, identifying new analogs with better properties is important. Synthesizing these new compounds is a logical next step, and is key to research groups that have a synthetic chemistry team or external collaborators. Generative machine learning models have become widely adopted to generate new molecules and explore molecular space, with the goal of discovering novel compounds with desires properties. These generative models have been composed from recurrent neural networks (RNNs), Variational Autoencoders (VAEs), and Generative Adversarial Networks (GANs) and are often combined with transfer learning or scoring of physicochemical properties to steer generative design. While these generative models have proven useful in generating new molecular libraries, often they are not capable of addressing a wide variety of potential problems, and often converge into similar molecular space when combined with a scoring function for desired properties. In addition, generated compounds are often not synthetically feasible, reducing their capabilities outside of virtual composition and limiting their usefulness in real-world scenarios. Here we introduce a suite of automated tools called MegaSyn representing 3 components: a new hill-climb algorithm which makes use of SMILES-based RNN generative models, analog generation software, and retrosynthetic analysis coupled with fragment analysis to score molecules for their synthetic feasibility. We now describe the development and testing of this suite of tools and propose how they might be used to optimize molecules or prioritize promising lead compounds using test case examples.