PaccMannRL: De novo generation of hit-like anticancer molecules from transcriptomic data via reinforcement learning

•A conditional generative model for de novo design of anticancer hit molecules is devised•Drug sensitivity and toxicity models steer the molecule design via reinforcement learning•Molecules are designed to target individual transcriptomic profiles of cell lines•Targeted, hit-like molecules are generated more frequently, even for unseen cell lines•In silico, the molecules exhibit similar physicochemical properties to real cancer drugs With the advent of deep generative models in computational chemistry, in-silico drug design is undergoing an unprecedented transformation. Although deep learning approaches have shown potential in generating compounds with desired chemical properties, they disregard the cellular environment of target diseases. Bridging systems biology and drug design, we present a reinforcement learning method for de novo molecular design from gene expression profiles. We construct a hybrid Variational Autoencoder that tailors molecules to target-specific transcriptomic profiles, using an anticancer drug sensitivity prediction model (PaccMann) as reward function. Without incorporating information about anticancer drugs, the molecule generation is biased toward compounds with high predicted efficacy against cell lines or cancer types. The generation can be further refined by subsidiary constraints such as toxicity. Our cancer-type-specific candidate drugs are similar to cancer drugs in drug-likeness, synthesizability, and solubility and frequently exhibit the highest structural similarity to compounds with known efficacy against these cancer types. With the advent of deep generative models in computational chemistry, in-silico drug design is undergoing an unprecedented transformation. Although deep learning approaches have shown potential in generating compounds with desired chemical properties, they disregard the cellular environment of target diseases. Bridging systems biology and drug design, we present a reinforcement learning method for de novo molecular design from gene expression profiles. We construct a hybrid Variational Autoencoder that tailors molecules to target-specific transcriptomic profiles, using an anticancer drug sensitivity prediction model (PaccMann) as reward function. Without incorporating information about anticancer drugs, the molecule generation is biased toward compounds with high predicted efficacy against cell lines or cancer types. The generation can be further refined by subsidiary constraints such as toxicity. Our cancer-type-specific candidate drugs are similar to cancer drugs in drug-likeness, synthesizability, and solubility and frequently exhibit the highest structural similarity to compounds with known efficacy against these cancer types. Eroom's law describes the observation that the productivity of the drug discovery pipeline, as measured by the number of FDA-approved drugs per billion US dollar invested, has been halved every 9 years since the 1950s (Scannell et al., 2012Scannell J.W. Blanckley A. Boldon H. Warrington B. Diagnosing the decline in pharmaceutical r&d efficiency.Nat. Rev. Drug Discov. 2012; 11: 191Crossref PubMed Scopus (1100) Google Scholar). Indeed, only a minimal portion of all synthesized drug candidates obtain market approval (less than 0.01%), with an estimated 10–15 years until market release and costs that range between one (Scannell et al., 2012Scannell J.W. Blanckley A. Boldon H. Warrington B. Diagnosing the decline in pharmaceutical r&d efficiency.Nat. Rev. Drug Discov. 2012; 11: 191Crossref PubMed Scopus (1100) Google Scholar) and three billion dollars per drug (DiMasi et al., 2016DiMasi J.A. Grabowski H.G. Hansen R.W. Innovation in the pharmaceutical industry: new estimates of r&d costs.J. Health Econ. 2016; 47: 20-33Crossref PubMed Scopus (1266) Google Scholar). This low efficiency has been attributed to the high dropout rate of candidate molecules in the early stages of the pipeline, highlighting the need for more accurate in silico and in vitro models that produce more potent candidate drugs. In addition to the initial wet-lab validations, the discovery pipeline involves a sequential process that builds upon high-throughput screenings, ADMET-assessments, and a lengthy phase of clinical trials. The costs of the experimental and clinical phase can be prohibitive and any solution that helps to reduce the number of required experimental assays can provide a competitive advantage and reduce time to market. The problem's linchpin is on how to improve the exploration and navigation through the chemical space that has been estimated to contain ∼1030–1060 drug-like molecules with bioactive properties (Polishchuk et al., 2013Polishchuk P.G. Madzhidov T.I. Varnek A. Estimation of the size of drug-like chemical space based on gdb-17 data.J. Comput. Aided Mol. Des. 2013; 27: 675-679Crossref PubMed Scopus (163) Google Scholar). Deep learning methods have recently gained popularity to aid drug discovery (Chen et al., 2018Chen H. Engkvist O. Wang Y. Olivecrona M. Blaschke T. The rise of deep learning in drug discovery.Drug Discov. Today. 2018; 23: 1241-1250Crossref PubMed Scopus (554) Google Scholar) and many have demonstrated the feasibility of in silico design of novel candidate compounds with desired chemical properties (Popova et al., 2018Popova M. Isayev O. Tropsha A. Deep reinforcement learning for de novo drug design.Sci. Adv. 2018; 4: eaap7885Crossref PubMed Scopus (311) Google Scholar; Gomez-Bombarelli et al., 2018Gomez-Bombarelli R. Wei J.N. Duvenaud D. Hernandez-Lobato J.M. Sánchez-Lengeling B. Sheberla D. Aguilera-Iparraguirre J. Hirzel T.D. Adams R.P. Aspuru-Guzik A. et al.Automatic chemical design using a data-driven continuous representation of molecules.ACS Cent. Sci. 2018; 4: 268-276Crossref PubMed Scopus (800) Google Scholar; You et al., 2018You J. Liu B. Ying Z. Pande V. Leskovec J. Graph convolutional policy network for goal-directed molecular graph generation.Adv. Neural Inf. Process. Syst. 2018; 31: 6410-6421Google Scholar). In all of these models, the generative process is controlled by a structurally driven evaluator (or critic) that biases the generation of a chemical to satisfy the required chemical structural properties. Although very effective in generating compounds with desired chemical properties, these methods disregard system-level information, e.g. about the cellular environment in which the drug is intended to act. However, the two main causes of the increasing attrition rate in drug design are a lack in efficacy against the specific disease of interest and off-target cytotoxicity (Wehling, 2009Wehling M. Assessing the translatability of drug projects: what needs to be scored to predict success? Nature reviews.Drug Discov. 2009; 8: 541-546Crossref Scopus (96) Google Scholar), calling to bridge systems biology closer with drug discovery. Related methodology has been used for protein-targeting de novo generation (Zhavoronkov et al., 2019Zhavoronkov A. Ivanenkov Y.A. Aliper A. Veselov M.S. Aladinskiy V.A. Aladinskaya A.V. Terentiev V.A. Polykovskiy D.A. Kuznetsov M.D. Asadulaev A. et al.Deep learning enables rapid identification of potent ddr1 kinase inhibitors.Nat. Biotechnol. 2019; 37: 1038-1040Crossref PubMed Scopus (240) Google Scholar; Aumentado-Armstrong, 2018Aumentado-Armstrong T. Latent Molecular Optimization for Targeted Therapeutic Design. arXiv, 2018: 1809.02032Google Scholar; Grechishnikova, 2021Grechishnikova D. Transformer neural network for protein-specific de novo drug generation as a machine translation problem.Sci. Rep. 2021; 11: 1-13Crossref PubMed Scopus (7) Google Scholar; Chenthamarakshan et al., 2020Chenthamarakshan V. Das P. Hoffman S.C. Strobelt H. Padhi I. Lim K.W. Hoover B. Manica M. Born J. Laino T. Mojsilovic A. Cogmol: target-specific and selective drug design for COVID-19 using deep generative models.Adv. Neural Inf. Process. Syst. 2020; 33PubMed Google Scholar; Skalic et al., 2019Skalic M. Sabbadin D. Sattarov B. Sciabola S. De Fabritiis G. From target to drug: generative modeling for the multimodal structure-based ligand design.Mol. Pharm. 2019; 16: 4282-4291Crossref PubMed Scopus (20) Google Scholar; Krishnan et al., 2021Krishnan S.R. Bung N. Bulusu G. Roy A. Accelerating de novo drug design against novel proteins using deep learning.J. Chem. Inf. Model. 2021; 61: 621-630Crossref PubMed Scopus (8) Google Scholar). These contributions attempt to utilize deep learning methods for de novo design of compounds to specifically target a protein that has been implicated in tumor proliferation or treatment response (e.g. gene-knockout study). For example, the study by Zhavoronkov et al., 2019Zhavoronkov A. Ivanenkov Y.A. Aliper A. Veselov M.S. Aladinskiy V.A. Aladinskaya A.V. Terentiev V.A. Polykovskiy D.A. Kuznetsov M.D. Asadulaev A. et al.Deep learning enables rapid identification of potent ddr1 kinase inhibitors.Nat. Biotechnol. 2019; 37: 1038-1040Crossref PubMed Scopus (240) Google Scholar curated and utilized, among others, patent data and several datasets about molecules (unspecific bioactive compounds, kinase inhibitors, DDR1 kinase inhibitors, molecules targeting non-kinase targets) specifically to develop DDR1 inhibitors. They synthesized and tested six drug candidates in cell assays. Two of them were found to be active, and one was even successfully validated in animal models. Envisioning a precision or even personalized medicine perspective, identifying protein targets is challenging, whereas sequencing and omics data are straightforward to gather. Very recently, Méndez-Lucio et al., 2020Méndez-Lucio O. Baillif B. Clevert D.-A. Rouquié D. Wichard J. De novo generation of hit-like molecules from gene expression signatures using artificial intelligence.Nat. Commun. 2020; 11: 1-10Crossref PubMed Scopus (53) Google Scholar proposed a method for the de novo design of molecules against desired targets, represented by the gene expression signatures of knocked-out (suspected) targets. Notably, 97% of all anticancer candidate drugs fail in clinical trials and never receive FDA approval, questioning the current approaches of target identification for the discovery of pharmaceuticals (Wong et al., 2019Wong C.H. Siah K.W. Lo A.W. Estimation of clinical trial success rates and related parameters.Biostatistics. 2019; 20: 273-286Crossref PubMed Scopus (347) Google Scholar). Taking ten drug-indication pairs from ongoing clinical trials, Lin et al., 2019Lin A. Giuliano C.J. Palladino A. John K.M. Abramowicz C. Yuan M.L. Sausville E.L. Lukow D.A. Liu L. Chait A.R. et al.Off-target toxicity is a common mechanism of action of cancer drugs undergoing clinical trials.Sci. Transl. Med. 2019; 11: eaaw8412Crossref PubMed Scopus (157) Google Scholar found that the proposed mechanism of action (MOA) for all of them were incorrect; knocking out the target genes did not ever hamper cancer fitness. Although the wrong target genes were identified through RNA interference with siRNA, seemingly silencing essential off-target genes, all drugs retained their anticancer effect through target-independent mechanisms. The fact that off-target cytotoxicity is a common MOA of anticancer drugs in clinical trials corroborates the need to to scrutinize current lead compound discovery strategies and calls to develop novel methodology with unconventional approaches. It is for this reason that we herein propose a novel framework to generate lead compound candidates solely based on a tumor's metabolic signature, as opposed to attempting to target a specific protein or incorporating information about potential targets directly into the design process. Here, we guide the learning process solely by transcriptomic profiles of cancer cells, which thus act as metabolic signatures. Transcriptomic data have been successfully used for de novo drug identification (Verbist et al., 2015Verbist B. Klambauer G. Vervoort L. Talloen W. Shkedy Z. Thas O. Bender A. Göhlmann H.W. Hochreiter S. Consortium Q. et al.Using transcriptomics to guide lead optimization in drug discovery projects: lessons learned from the qstar project.Drug Discov. Today. 2015; 20: 505-513Crossref PubMed Scopus (56) Google Scholar; De Wolf et al., 2018De Wolf H. Cougnaud L. Van Hoorde K. De Bondt A. Wegner J.K. Ceulemans H. Göhlmann H. High-throughput gene expression profiles to define drug similarity and predict compound activity.Assay Drug Dev. Tech. 2018; 16: 162-176Crossref PubMed Scopus (13) Google Scholar) and has been advocated for a pivotal role in the future of drug discovery (Dopazo, 2014Dopazo J. Genomics and transcriptomics in drug discovery.Drug Discov. Today. 2014; 19: 126-132Crossref PubMed Scopus (41) Google Scholar). Related work has addressed the generation of anticancer candidate drugs by conditionally sampling from an IC50 vector (Joo et al., 2020Joo S. Kim M.S. Yang J. Park J. Generative model for proposing drug candidates satisfying anticancer properties using a conditional variational autoencoder.ACS Omega. 2020; 5: 18642-18650Crossref PubMed Scopus (5) Google Scholar). We present a novel framework for molecule generation based on deep generative models and reinforcement learning that, for the first time, enables the generation of molecules while taking into account the disease context encoded in the form of gene expression profile (GEP) data (for a graphical illustration see Figure 1A). Our framework is depicted in Figure 1B and consists of a conditional molecule generator (embodied by two separate Variational Autoencoders) and a critic module that evaluates the efficacy of proposed compounds on the target profile (see Figure 1D). The training procedure is split into two stages. In the first stage, the models are trained independently; one VAE is trained on gene expression data (in the following called profile VAE or just PVAE) from TCGA (Weinstein et al., 2013Weinstein J.N. Collisson E.A. Mills G.B. Shaw K.R.M. Ozenberger B.A. Ellrott K. Shmulevich I. Sander C. Stuart J.M. Network C.G.A.R. et al.The cancer genome atlas pan-cancer analysis project.Nat. Genet. 2013; 45: 1113Crossref PubMed Scopus (3265) Google Scholar), and another VAE (in the following called SMILES VAE or just SVAE) is trained on bioactive small molecules from ChEMBL (Bento et al., 2013Bento A.P. Gaulton A. Hersey A. Bellis L.J. Chambers J. Davies M. Krüger F.A. Light Y. Mak L. McGlinchey S. et al.The ChEMBL bioactivity database: an update.Nucleic Acids Res. 2013; 42: D1083-D1090Crossref PubMed Scopus (973) Google Scholar) (see Figure 1C). As a critic, we use PaccMann, a multimodal drug sensitivity prediction model developed and validated in our previous work (Manica et al., 2019Manica M. Oskooei A. Born J. Subramanian V. Saez-Rodriguez J. Rodriguez Martinez M. Toward explainable anticancer compound sensitivity prediction via multimodal attention-based convolutional encoders.Mol. Pharm. 2019; 16: 4797-4806Crossref PubMed Scopus (30) Google Scholar; Cadow et al., 2020Cadow J. Born J. Manica M. Oskooei A. Rodríguez Martínez M. Paccmann: a web service for interpretable anticancer compound sensitivity prediction.Nucleic Acids Res. 2020; 48: W502-W508Crossref PubMed Google Scholar). In the second stage, the encoder of the profile VAE is combined with the decoder of the molecule VAE and exposed to a joint retraining that is optimized using a policy gradient regime with a reward coming from the critic module. The goal of the optimization is to tune the generative model such that it generates (novel) compounds that have maximal efficacy against a given biomolecular profile that is characteristic for a cancer site, a patient subgroup, or even an individual. By efficacy, we refer to predict cellular IC50 (i.e. the micromolar concentration necessary to inhibit 50% of the cells) as opposed to e.g. enzymatic IC50. This efficacy is a joint property of a drug-cell-pair, as treatment response to a compound heavily varies depending on the tumor's genomic and transcriptomic makeup (Geeleher et al., 2016Geeleher P. Cox N.J. Huang R.S. Cancer biomarker discovery is improved by accounting for variability in general levels of drug sensitivity in pre-clinical models.Genome Biol. 2016; 17: 190Crossref PubMed Scopus (24) Google Scholar). In this work, we focus on profile-specific compound generation and optimize the generator with IC50 as critic, but we would like to note that our framework can be extended to more complex reward functions and include a case study on the concurrent optimization of (predicted) drug efficacy and toxicity. In the first phase of training, the two components depicted in Figure 1C were trained independently. The profile VAE consisted of a set of stacked dense layers and was trained as a denoizing VAE to enhance generalization abilities. The purpose of the PVAE was to find a lower dimensional representation of the cell profiles that maintains structural similarity and later allows a fusion with the latent representation of molecules. The encoder of the PVAE learned to meaningfully embed gene expression profiles (bulk RNA-Seq from TCGA (Weinstein et al., 2013Weinstein J.N. Collisson E.A. Mills G.B. Shaw K.R.M. Ozenberger B.A. Ellrott K. Shmulevich I. Sander C. Stuart J.M. Network C.G.A.R. et al.The cancer genome atlas pan-cancer analysis project.Nat. Genet. 2013; 45: 1113Crossref PubMed Scopus (3265) Google Scholar)) into a latent space, such that the decoder could reconstruct the profiles, but also generate novel, seemingly realistic gene expression profiles (GEP). In alignment with the reported consensus between transcriptomic data in TCGA and cancer cell line databases (Ghandi et al., 2019Ghandi M. Huang F.W. Jané-Valbuena J. Kryukov G.V. Lo C.C. McDonald E.R. Barretina J. Gelfand E.T. Bielski C.M. Li H. et al.Next-generation characterization of the cancer cell line encyclopedia.Nature. 2019; 569: 503Crossref PubMed Scopus (658) Google Scholar), we found the distributions of GEPs in GDSC (Yang et al., 2012Yang W. Soares J. Greninger P. Edelman E.J. Lightfoot H. Forbes S. Bindal N. Beare D. Smith J.A. Thompson IR et al.Genomics of drug sensitivity in cancer (gdsc): a resource for therapeutic biomarker discovery in cancer cells.Nucleic Acids Res. 2012; 41: D955-D961Crossref PubMed Scopus (911) Google Scholar) and TCGA to be sufficiently similar to justify our decision to perform the PVAE pretraining on ∼10k TCGA samples, whereas the reinforcement learning (RL) optimization was performed on GDSC cell lines (compare Figure S2 in supplementary material S2). The SMILES VAE was pretrained for 10 epochs on ∼1.4 million structures from ChEMBL (Bento et al., 2013Bento A.P. Gaulton A. Hersey A. Bellis L.J. Chambers J. Davies M. Krüger F.A. Light Y. Mak L. McGlinchey S. et al.The ChEMBL bioactivity database: an update.Nucleic Acids Res. 2013; 42: D1083-D1090Crossref PubMed Scopus (973) Google Scholar) (for details see transparent methods). Both encoder and decoder consisted of stack-augmented gated-recurrent units as used in Popova et al., 2018Popova M. Isayev O. Tropsha A. Deep reinforcement learning for de novo drug design.Sci. Adv. 2018; 4: eaap7885Crossref PubMed Scopus (311) Google Scholar. The purpose of the SVAE was to learn the syntax of SMILES and general semantics about bioactive compounds. The novelty and diversity of the generated molecules were validated by sampling 10,000 molecules through decoding random points from the latent space. About 96.2% of the 10,000 generated molecules were valid molecular structures (surpassing the results of Popova et al., 2018Popova M. Isayev O. Tropsha A. Deep reinforcement learning for de novo drug design.Sci. Adv. 2018; 4: eaap7885Crossref PubMed Scopus (311) Google Scholar who used the same stack-augmented GRUs and reported 95% SMILES validity). Moreover, 99.72% of the valid molecules were unique across the 10,000 generations, and the novelty was 1, i.e. none of the generated compounds was present in the training dataset. Comparing the Tanimoto similarity (Tanimoto, 1958Tanimoto T.T. Elementary Mathematical Theory of Classification and Prediction. IBM Internal Report, 1958Google Scholar) of the ECFP molecular fingerprints (Rogers and Hahn, 2010Rogers D. Hahn M. Extended-connectivity fingerprints.J. Chem. Inf. Model. 2010; 50: 742-754Crossref PubMed Scopus (2360) Google Scholar) of 1,000 generated molecules with the training and test data from ChEMBL, we found that the vast majority had a Tanimoto similarity (τ), between 0.2 and 0.6 (on average 0.41 ± 0.1 for training and 0.38 ± 0.08 for testing molecules), suggesting that our model learned to propose novel molecular structures from the chemical space. In addition, an interactive visualization of the chemical space with Faerun (Probst and Reymond, 2018Probst D. Reymond J.-L. Fun: a framework for interactive visualizations of large, high-dimensional datasets on the web.Bioinformatics. 2018; 34: 1433-1435Crossref PubMed Scopus (22) Google Scholar) shows the TMAP (Probst and Reymond, 2020Probst D. Reymond J.-L. Visualization of very large high-dimensional data sets as minimum spanning trees.J. Cheminform. 2020; 12: 1-13Crossref PubMed Scopus (46) Google Scholar) (i.e., an algorithm to represent high-dimensional data as minimum spanning trees) of ChEMBL and generated compounds through the TMAP algorithm. The generated molecules mix well with the training molecules into the chemical space. For detailed results of both the PVAE and the SVAE, see the supplementary material S3. Here, we present the results of our molecule generator conditioned on gene expression profiles of cancer subtypes. As a proof of concept, we show results for cancer in four different sites: breast (carcinoma), lung (carcinoma), prostate (carcinoma), and autonomic ganglia (neuroblastoma). The conditional generator was initialized as the SVAE, i.e. sampling from the unbiased generator yielded random molecules from the chemical space as learned from the ChEMBL data. For the evaluation, all generated compounds with a predicted IC50 value below 1 μM (i.e. pIC50 > 6) were considered to be effective. Moreover, within each cancer type (or site), 80% of the cell lines (breast: 50, lung: 169, prostate: 7, autonomic ganglia: 56) were considered as training cell lines and used to optimize the parameters of the conditional generator. We observed that over time the generator learned to produce more molecules with high predicted efficacy according to the critic. To test both the generalization abilities and whether the generator actually utilized the omics-profile for the generation, we used the remaining 20% of cell lines to verify whether conditioning the generator on unseen cell lines of the same site also leads to compounds with low IC50. As presented in Figure 2 (left column), our model learned to produce compounds with lower IC50 values, for unseen cell lines from the given cancer site. In other words, the IC50 distribution of candidate compounds proposed by the generative model were successfully shifted toward higher efficacy (lower IC50). The baseline model corresponds to the pretrained SVAE from which n = 500 molecules were randomly sampled. In all four cases, a significant portion (between 16% and 57%) of molecules generated from the optimized model were assigned an IC50 value below 1μM, whereas only 2%–5% of the candidates generated by the baseline model (i.e. the SVAE) were classified as effective. Moreover, in all cases the generator maintained an almost equal SMILES validity (84%–93%) compared with the baseline, much higher than what Méndez-Lucio et al., 2020Méndez-Lucio O. Baillif B. Clevert D.-A. Rouquié D. Wichard J. De novo generation of hit-like molecules from gene expression signatures using artificial intelligence.Nat. Commun. 2020; 11: 1-10Crossref PubMed Scopus (53) Google Scholar reported based on gene expression (8%–9%). The second column of Figure 2 shows generated molecules that are predicted to be effective against unseen cell lines from the respective cancer site. As opposed to the personalized regime in the second column, the third column of Figure 2 showcases a precision medicine regime. Here, novel molecules were designed specifically for each cancer site, i.e. a single, characteristic GEP. In all cases, the model generated compounds that exhibited high efficacy against the average cellular profile of the target site while maintaining efficacy against the majority of individual cell lines for that site. We have thus formulated a novel problem, namely how to drive a molecular generative model to produce molecules with low predicted efficacy against an omics profile of interest. To the best of our knowledge, this problem has not been tackled before, exacerbating a comparison to other works. In the most similar work, Méndez-Lucio et al., 2020Méndez-Lucio O. Baillif B. Clevert D.-A. Rouquié D. Wichard J. De novo generation of hit-like molecules from gene expression signatures using artificial intelligence.Nat. Commun. 2020; 11: 1-10Crossref PubMed Scopus (53) Google Scholar proposed a model that produces hit-like molecules to induce a desired gene expression profile. For a more quantitative assessment, the last column of Figure 2 compares the four cancer-type-specific candidate compounds with one of their top-3 neighbors using the Tanimoto similarity score, τ, from several hundreds of existing anticancer compounds. It is well known that the Tanimoto similarity across compounds is highly correlated with their induced sensitivity patterns on cancer cell lines (Shivakumar and Krauthammer, 2009Shivakumar P. Krauthammer M. Structural similarity assessment for drug sensitivity prediction in cancer.in: BMC Bioinformatics. volume 10. Springer, 2009: S17Google Scholar). The candidate compound proposed against breast cancer (Figure 2 second row, third column) resembles a collection of fused sugarlike moieties and has doxorubicin, a commonly used chemotherapeutical against breast cancer (Lao et al., 2013Lao J. Madani J. Puértolas T. Álvarez M. Hernández A. Pazo-Cid R. Artal Á. Antón Torres A. Liposomal doxorubicin in the treatment of breast cancer patients: a review.J. Drug Deliv. 2013; 2013: 456409Crossref PubMed Google Scholar), as one of the top-3 nearest neighbors. The generated compound against lung cancer (Figure 2, third row, third column) presents similarities to embelin, an existing anticancer compound from the GDSC database. Comparing the two structures, it is evident that the generated compound and embelin share a long carbon chain and a single six-membered fully carbonic ring. Embelin was tested against 965 cell lines from GDSC/CCLE from which the highest reported efficacy is against a lung cell line (NT2-D1). Embelin is also known to be the only known non-peptide inhibitor of XIAP (Poojari, 2014Poojari R. Embelin–a drug of antiquity: shifting the paradigm towards modern medicine.Expert Opin. Investig. Drugs. 2014; 23: 427-444Crossref PubMed Scopus (44) Google Scholar), a protein that plays an important role in lung cancer development (Cheng et al., 2010Cheng Y.-J. Jiang H.-S. Hsu S.-L. Lin L.-C. Wu C.-L. Ghanta V.K. Hsueh C.-M. Xiap-mediated protection of h460 lung cancer cells against cisplatin.Eur. J. Pharmacol. 2010; 627: 75-84Crossref PubMed Scopus (35) Google Scholar). The closest neighbor of the prostate-specific generated compound (Figure 2 fourth row, third column) is vorapaxar. Its efficacy is highest against a prostate cancer cell line (DU_145) according to GDSC/CCLE. Vorapaxar is an antagonist of a protease-activated receptor (PAR-1) that is known to be overexpressed in various types of cancer, including prostate (Zhang et al., 2009Zhang X. Wang W. True L.D. Vessella R.L. Takayama T.K. Protease-activated receptor-1 is upregulated in reactive stroma of primary prostate cancer and bone metastasis.Prostate. 2009; 69: 727-736Crossref PubMed Scopus (23) Google Scholar). Lastly, the third closest neighbor of the generated compound against neuroblastoma (Figure 2 first row, third column) is fulvestrant, an antagonist/modulator of ERα that has recently been proposed as a novel anticancer agent for neuroblastoma (Gorska et al., 2016Gorska M. Kuban-Jankowska A. Milczarek R. Wozniak M. Nitro-oxidative stress is involved in anticancer activity of 17β-estradiol derivative in neuroblastoma cells.Anticancer Res. 2016; 36: 1693-1698PubMed Google Scholar). The predicted pIC50 activity profile of fulvestrant and our candidate drug are highly correlated across all cell lines (ρ = 0.88), indicating that they may exhibit similar pharmacological properties. Similar observations are made for the lung and prostate cancer candidates and their neighbors embelin and vorapaxar (ρ = 0.55 and ρ = 0.69). To summarize, for all four investigated cancer types, the proposed compounds showed the highest structural similarity to anticancer drugs that are, for each specific cancer type, either (1) already FDA approved (breast), (2) known inhibitors of relevant targets (lung, prostate), or (3) have been advocated for (neuroblastoma). This result is remarkable, especially as the generator was never exposed to any anticancer compounds. Indeed, only the critic had seen two out of the four compounds during training, highlighting that the generator has learned some structural characteristics that make a compound efficacious against a particular cancer type, according to our critique. In the above comparisons, the search space was restricted to compounds with known anticancer properties. To investigate whether the proposed compounds generally had a higher similarity to drugs associated with cancer, we carried out a comparison with compounds from a broader pool of chemicals, namely ChEMBL (Bento et al., 2013Bento A.P. Gaulton A. Hersey A. Bellis L.J. Chambers J. Davies M. Krüger F.A. Light Y. Mak L. McGlinchey S. et al.The ChEMBL bioactivity database: an update.Nucleic Acids Res. 2013; 42: D1083-D1090Crossref PubMed Scopus