Chemoinformatics and artificial intelligence colloquium: progress and challenges in developing bioactive compounds.

We report the main conclusions of the first Chemoinformatics and Artificial Intelligence Colloquium, Mexico City, June 15-17, 2022. Fifteen lectures were presented during a virtual public event with speakers from industry, academia, and non-for-profit organizations. Twelve hundred and ninety students and academics from more than 60 countries. During the meeting, applications, challenges, and opportunities in drug discovery, de novo drug design, ADME-Tox (absorption, distribution, metabolism, excretion and toxicity) property predictions, organic chemistry, peptides, and antibiotic resistance were discussed. The program along with the recordings of all sessions are freely available at https://www.difacquim.com/english/events/2022-colloquium/ .

Exploring the Use of Compound-Induced Transcriptomic Data Generated From Cell Lines to Predict Compound Activity Toward Molecular Targets.

Pharmaceutical or phytopharmaceutical molecules rely on the interaction with one or more specific molecular targets to induce their anticipated biological responses. Nonetheless, these compounds are also prone to interact with many other non-intended biological targets, also known as off-targets. Unfortunately, off-target identification is difficult and expensive. Consequently, QSAR models predicting the activity on a target have gained importance in drug discovery or in the de-risking of chemicals. However, a restricted number of targets are well characterized and hold enough data to build such in silico models. A good alternative to individual target evaluations is to use integrative evaluations such as transcriptomics obtained from compound-induced gene expression measurements derived from cell cultures. The advantage of these particular experiments is to capture the consequences of the interaction of compounds on many possible molecular targets and biological pathways, without having any constraints concerning the chemical space. In this work, we assessed the value of a large public dataset of compound-induced transcriptomic data, to predict compound activity on a selection of 69 molecular targets. We compared such descriptors with other QSAR descriptors, namely the Morgan fingerprints (similar to extended-connectivity fingerprints). Depending on the target, active compounds could show similar signatures in one or multiple cell lines, whether these active compounds shared similar or different chemical structures. Random forest models using gene expression signatures were able to perform similarly or better than counterpart models built with Morgan fingerprints for 25% of the target prediction tasks. These performances occurred mostly using signatures produced in cell lines showing similar signatures for active compounds toward the considered target. We show that compound-induced transcriptomic data could represent a great opportunity for target prediction, allowing to overcome the chemical space limitation of QSAR models.

De novo generation of hit-like molecules from gene expression signatures using artificial intelligence.

Finding new molecules with a desired biological activity is an extremely difficult task. In this context, artificial intelligence and generative models have been used for molecular de novo design and compound optimization. Herein, we report a generative model that bridges systems biology and molecular design, conditioning a generative adversarial network with transcriptomic data. By doing so, we can automatically design molecules that have a high probability to induce a desired transcriptomic profile. As long as the gene expression signature of the desired state is provided, this model is able to design active-like molecules for desired targets without any previous target annotation of the training compounds. Molecules designed by this model are more similar to active compounds than the ones identified by similarity of gene expression signatures. Overall, this method represents an alternative approach to bridge chemistry and biology in the long and difficult road of drug discovery.

Lysine harvesting is an antioxidant strategy and triggers underground polyamine metabolism.

Both single and multicellular organisms depend on anti-stress mechanisms that enable them to deal with sudden changes in the environment, including exposure to heat and oxidants. Central to the stress response are dynamic changes in metabolism, such as the transition from the glycolysis to the pentose phosphate pathway-a conserved first-line response to oxidative insults1,2. Here we report a second metabolic adaptation that protects microbial cells in stress situations. The role of the yeast polyamine transporter Tpo1p3-5 in maintaining oxidant resistance is unknown6. However, a proteomic time-course experiment suggests a link to lysine metabolism. We reveal a connection between polyamine and lysine metabolism during stress situations, in the form of a promiscuous enzymatic reaction in which the first enzyme of the polyamine pathway, Spe1p, decarboxylates lysine and forms an alternative polyamine, cadaverine. The reaction proceeds in the presence of extracellular lysine, which is taken up by cells to reach concentrations up to one hundred times higher than those required for growth. Such extensive harvest is not observed for the other amino acids, is dependent on the polyamine pathway and triggers a reprogramming of redox metabolism. As a result, NADPH-which would otherwise be required for lysine biosynthesis-is channelled into glutathione metabolism, leading to a large increase in glutathione concentrations, lower levels of reactive oxygen species and increased oxidant tolerance. Our results show that nutrient uptake occurs not only to enable cell growth, but when the nutrient availability is favourable it also enables cells to reconfigure their metabolism to preventatively mount stress protection.

Data-driven approaches used for compound library design, hit triage and bioactivity modeling in high-throughput screening.

High-throughput screening (HTS) campaigns are routinely performed in pharmaceutical companies to explore activity profiles of chemical libraries for the identification of promising candidates for further investigation. With the aim of improving hit rates in these campaigns, data-driven approaches have been used to design relevant compound screening collections, enable effective hit triage and perform activity modeling for compound prioritization. Remarkable progress has been made in the activity modeling area since the recent introduction of large-scale bioactivity-based compound similarity metrics. This is evidenced by increased hit rates in iterative screening strategies and novel insights into compound mode of action obtained through activity modeling. Here, we provide an overview of the developments in data-driven approaches, elaborate on novel activity modeling techniques and screening paradigms explored and outline their significance in HTS.

Towards understanding polyol additive effects on the pH shift-induced aggregation of a monoclonal antibody using high throughput screening and quantitative structure-activity modeling.

pH shift-induced aggregation is frequently observed in downstream processing of monoclonal antibodies and has been shown to depend on solvent composition. To quantify the stabilizing effect of polyol additives against aggregation, we determined aggregation rate constants in the presence of a set of 14 compounds. Rate constants were then correlated with molecular descriptors in a quantitative structure activity relationship (QSAR) approach. The molecular size, volume, the charge, number of hydrogen acceptors, the stereochemistry and hydrophobicity of the compounds were identified as important descriptors. Generally larger compounds with a balanced surface polarity tend to inhibit aggregation better while hydrophobicity plays an important role at the nucleation phase, with hydrophobic compounds being more potent at inhibiting aggregation.

Activity Landscape Plotter: A Web-Based Application for the Analysis of Structure-Activity Relationships.

Activity landscape modeling is a powerful method for the quantitative analysis of structure-activity relationships. This cheminformatics area is in continuous growth, and several quantitative and visual approaches are constantly being developed. However, these approaches often fall into disuse due to their limited access. Herein, we present Activity Landscape Plotter as the first freely available web-based tool to automatically analyze structure-activity relationships of compound data sets. Based on the concept of activity landscape modeling, the online service performs pairwise structure and activity relationships from an input data set supplied by the user. For visual analysis, Activity Landscape Plotter generates Structure-Activity Similarity and Dual-Activity Difference maps. The user can interactively navigate through the maps and export all the pairwise structure-activity information as comma delimited files. Activity Landscape Plotter is freely accessible at https://unam-shiny-difacquim.shinyapps.io/ActLSmaps /.

The many roles of molecular complexity in drug discovery.

Molecular complexity is becoming a crucial concept in drug discovery. It has been associated with target selectivity, success in progressing into clinical development and compound safety, among other factors. Multiple metrics have been developed to quantify molecular complexity and explore complexity-property relationships. However, there is no general agreement regarding how to measure this molecular feature. Herein, we have surveyed the many roles of molecular complexity in drug discovery discussing in a critical manner different quantification methods. Through the analysis of various reference compound databases, common pitfalls and workarounds of the quantification of molecular complexity are discussed.

Synthesis, Screening and in silico Simulations of Anti-Parasitic Propamidine/Benzimidazole Derivatives.

BACKGROUND: We designed hybrid molecules between propamidine and benzimidazole in order to retain the antiprotozoal action, but decreasing the toxic effect of the molecule. OBJECTIVE: Design and prepare 12 hybrids for testing their antiparasitic effect over three protozoa: Giardia intestinalis, Trichomonas vaginalis and Leishmania mexicana, as well as conduct several in silico simulations such as toxicological profile, molecular docking and molecular dynamics in order to understand their potential mode of action. METHODS: Hybrids 1-3, 6-9 and 12 were obtained using a chemical pathway previously reported. Compounds 4, 5, 10 and 11 were prepared using a one-pot reduction-cyclization reaction. The in vitro antiparasitic and cytotoxic activities of these compounds were conducted. It was calculated several properties such as toxicity, PK behavior, as well as docking studies and molecular dynamics of the most active compound performed in a DNA sequence dodecamer in comparison with propamidine. RESULTS: Compound 2 was 183, 127 and 202 times more active against G. intestinalis than metronidazole, pentamidine and propamidine. It was eleven times more active than pentamidine against L. mexicana. This compound showed low in vitro mammalian cytotoxicity. Molecular simulations showed a stable complex 2-DNA that occurred in the minor groove, analogous to propamidine-DNA complex. CONCLUSION: Compound 2, exhibited the higher bioactivity, especially towards G. intestinalis and L. mexicana. This study demonstrated that the replacement of benzimidazole scaffold instead of toxic amidine group in propamidine, results in an enhancement of antiprotozoal bioactivity. The preliminary molecular dynamics simulation suggests that the ligand-DNA complex is stable.

Diversity selection, screening and quantitative structure-activity relationships of osmolyte-like additive effects on the thermal stability of a monoclonal antibody.

Solvents used for therapeutic proteins in downstream processing and in formulations often contain stabilizing additives that inhibit denaturation and aggregation. Such additives are mostly selected based on their positive effect on thermal stability of the protein, and are often derived from naturally occuring osmolytes. To better understand the structural basis underlying the effect of additives, we selected a diverse library of compounds comprising 79 compounds of the polyol, amino acid and methylamine chemical classes and determined the effect of each compound on thermal stability of a monoclonal antibody as a function of compound concentration. Thermal stabilization of the antibody was influenced by solution pH. Quantitative structure-activity relationships (QSAR) were derived by partial least squares regression for individual compound classes and globally. The global model suggests that ligands with a phenyl ring will decrease the Tm, while highly soluble, polar compounds with at least two hydrogen bond donors will increase the Tm. This approach may be beneficial for further studies on the influence of other solution conditions like ionic strength and buffer species on additive-mediated protein stabilization.

Chemoinformatic expedition of the chemical space of fungal products.

AIM: Fungi are valuable resources for bioactive secondary metabolites. However, the chemical space of fungal secondary metabolites has been studied only on a limited basis. Herein, we report a comprehensive chemoinformatic analysis of a unique set of 207 fungal metabolites isolated and characterized in a USA National Cancer Institute funded drug discovery project. RESULTS: Comparison of the molecular complexity of the 207 fungal metabolites with approved anticancer and nonanticancer drugs, compounds in clinical studies, general screening compounds and molecules Generally Recognized as Safe revealed that fungal metabolites have high degree of complexity. Molecular fingerprints showed that fungal metabolites are as structurally diverse as other natural products and have, in general, drug-like physicochemical properties. CONCLUSION: Fungal products represent promising candidates to expand the medicinally relevant chemical space. This work is a significant expansion of an analysis reported years ago for a smaller set of compounds (less than half of the ones included in the present work) from filamentous fungi using different structural properties.

Statistical correlation of nonconservative substitutions of HIV gp41 variable amino acid residues with the R5X4 HIV-1 phenotype.

BACKGROUND: The interaction of the envelope glycoprotein of HIV-1 (gp120/gp41) with coreceptor molecules has important implications for specific cellular targeting and pathogenesis. Experimental and theoretical evidences have shown a role for gp41 in coreceptor tropism, although there is no consensus about the positions involved. Here we analyze the association of physicochemical properties of gp41 amino acid residues with viral tropism (X4, R5, and R5X4) using a large set of HIV-1 sequences. Under the assumption that conserved regions define the complex structural features essential for protein function, we focused our search only on amino acids in the gp41 variable regions. METHODS: Gp41 amino acid sequences of 2823 HIV-1 strains from all clades with known coreceptor tropism were retrieved from Los Alamos HIV Database. Consensus sequences were constructed for homologous sequences (those obtained from the same patient and having the same tropism) in order to avoid bias due to sequence overrepresentation, and the variability (entropy) per site was determined. Comparisons of hydropathy index (HI) and charge (Q) of amino acid residues at highly variable positions between coreceptor groups were performed using two non-parametrical tests and Benjamini-Hochberg correction. Pearson's correlation analysis was performed to determine covariance of HI and Q values. RESULTS: Calculation of variability per site rendered 58 highly variable amino acid positions. Of these, statistical analysis rendered significantly different HI or Q only for the R5 vs. R5X4 comparison at twelve positions: 535, 602, 619, 636, 640, 641, 658, 662, 667, 723, 756 and 841. The largest differences in particular amino acid frequencies between coreceptor groups were found at 619, 636, 640, 641, 662, 723 and 756. A hydrophobic tendency of residues 619, 640, 641, 723 and 756, along with a hydrophilic/charged tendency at residues 636 and 662 was observed in R5X4 with respect to R5 sequences. HI of position 640 covariated with that of 602, 619, 636, 662, and 756. CONCLUSIONS: Variability and significant correlations of physicochemical properties with viral phenotype suggest that substitutions at residues in the loop (602 and 619), the HR2 (636, 640, 641, 662), and the C-terminal tail (723, 756) of gp41 may contribute to phenotype of R5X4 strains.

ARWAR: A network approach for predicting Adverse Drug Reactions.

Predicting novel drug side-effects, or Adverse Drug Reactions (ADRs), plays an important role in the drug discovery process. Existing methods consider mainly the chemical and biological characteristics of each drug individually, thereby neglecting information hidden in the relationships among drugs. Complementary to the existing individual methods, in this paper, we propose a novel network approach for ADR prediction that is called Augmented Random-WAlk with Restarts (ARWAR). ARWAR, first, applies an existing method to build a network of highly related drugs. Then, it augments the original drug network by adding new nodes and new edges to the network and finally, it applies Random Walks with Restarts to predict novel ADRs. Empirical results show that the ARWAR method presented here outperforms the existing network approach by 20% with respect to average Fmeasure. Furthermore, ARWAR is capable of generating novel hypotheses about drugs with respect to novel and biologically meaningful ADR.

Activity and property landscape modeling is at the interface of chemoinformatics and medicinal chemistry.

Property landscape modeling (PLM) methods are at the interface of experimental sciences and computational chemistry. PLM are becoming a common strategy to describe systematically structure-property relationships of datasets. Thus far, PLM have been used mainly in medicinal chemistry and drug discovery. Herein, we survey advances on key topics on PLM with emphasis on questions often raised regarding the outcomes of the property landscape studies. We also emphasize on concepts of PLM that are being extended to other experimental areas beyond drug discovery. Topics discussed in this paper include applications of PLM to further characterize protein-ligand interactions, the utility of PLM as a quantitative and descriptive approach, and the statistical validation of property cliffs.

Activity cliffs and activity cliff generators based on chemotype-related activity landscapes.

Activity cliffs have large impact in drug discovery; therefore, their detection and quantification are of major importance. This work introduces the metric activity cliff enrichment factor and expands the previously reported activity cliff generator concept by adding chemotype information to representations of the activity landscape. To exemplify these concepts, three molecular databases with multiple biological activities were characterized. Compounds in each database were grouped into chemotype classes. Then, pairwise comparisons of structure similarities and activity differences were calculated for each compound and used to construct chemotype-based structure-activity similarity (SAS) maps. Different landscape distributions among four major regions of the SAS maps were observed for different subsets of molecules grouped in chemotypes. Based on this observation, the activity cliff enrichment factor was calculated to numerically detect chemotypes enriched in activity cliffs. Several chemotype classes were detected having major proportion of activity cliffs than the entire database. In addition, some chemotype classes comprising compounds with smooth structure activity relationships (SAR) were detected. Finally, the activity cliff generator concept was applied to compounds grouped in chemotypes to extract valuable SAR information.

Analyzing multitarget activity landscapes using protein-ligand interaction fingerprints: interaction cliffs.

Activity landscape modeling is mostly a descriptive technique that allows rationalizing continuous and discontinuous SARs. Nevertheless, the interpretation of some landscape features, especially of activity cliffs, is not straightforward. As the nature of activity cliffs depends on the ligand and the target, information regarding both should be included in the analysis. A specific way to include this information is using protein-ligand interaction fingerprints (IFPs). In this paper we report the activity landscape modeling of 507 ligand-kinase complexes (from the KLIFS database) including IFP, which facilitates the analysis and interpretation of activity cliffs. Here we introduce the structure-activity-interaction similarity (SAIS) maps that incorporate information on ligand-target contact similarity. We also introduce the concept of interaction cliffs defined as ligand-target complexes with high structural and interaction similarity but have a large potency difference of the ligands. Moreover, the information retrieved regarding the specific interaction allowed the identification of activity cliff hot spots, which help to rationalize activity cliffs from the target point of view. In general, the information provided by IFPs provides a structure-based understanding of some activity landscape features. This paper shows examples of analyses that can be carried out when IFPs are added to the activity landscape model.

Discovery and development of DNA methyltransferase inhibitors using in silico approaches.

Multiple strategies have evolved during the past few years to advance epigenetic compounds targeting DNA methyltransferases (DNMTs). Significant progress has been made in HTS, lead optimization and determination of 3D structures of DNMTs. In light of the emerging concept of epi-informatics, computational approaches are employed to accelerate the development of DNMT inhibitors helping to screen chemical databases, mine the DNMT-relevant chemical space, uncover SAR and design focused libraries. Computational methods also synergize with natural-product-based drug discovery and drug repurposing. Herein, we survey the latest developments of in silico approaches to advance epigenetic drug and probe discovery targeting DNMTs.

The interplay between molecular modeling and chemoinformatics to characterize protein-ligand and protein-protein interactions landscapes for drug discovery.

Protein-ligand and protein-protein interactions play a fundamental role in drug discovery. A number of computational approaches have been developed to characterize and use the knowledge of such interactions that can lead to drug candidates and eventually compounds in the clinic. With the increasing structural information of protein-ligand and protein-protein complexes, the combination of molecular modeling and chemoinformatics approaches are often required for the efficient analysis of a large number of such complexes. In this chapter, we review the progress on the developments of in silico approaches that are at the interface between molecular modeling and chemoinformatics. Although the list of methods and applications is not exhaustive, we aim to cover representative cases with a special emphasis on interaction fingerprints and their applications to identify "hot spots." We also elaborate on proteochemometric modeling and the emerging concept of activity landscape, structure-based interpretation of activity cliffs and structure-protein-ligand interaction relationships. Target-ligand relationships are discussed in the context of chemogenomics data sets.

Modelling ligand selectivity of serine proteases using integrative proteochemometric approaches improves model performance and allows the multi-target dependent interpretation of features.

Serine proteases, implicated in important physiological functions, have a high intra-family similarity, which leads to unwanted off-target effects of inhibitors with insufficient selectivity. However, the availability of sequence and structure data has now made it possible to develop approaches to design pharmacological agents that can discriminate successfully between their related binding sites. In this study, we have quantified the relationship between 12,625 distinct protease inhibitors and their bioactivity against 67 targets of the serine protease family (20,213 data points) in an integrative manner, using proteochemometric modelling (PCM). The benchmarking of 21 different target descriptors motivated the usage of specific binding pocket amino acid descriptors, which helped in the identification of active site residues and selective compound chemotypes affecting compound affinity and selectivity. PCM models performed better than alternative approaches (models trained using exclusively compound descriptors on all available data, QSAR) employed for comparison with R(2)/RMSE values of 0.64 ± 0.23/0.66 ± 0.20 vs. 0.35 ± 0.27/1.05 ± 0.27 log units, respectively. Moreover, the interpretation of the PCM model singled out various chemical substructures responsible for bioactivity and selectivity towards particular proteases (thrombin, trypsin and coagulation factor 10) in agreement with the literature. For instance, absence of a tertiary sulphonamide was identified to be responsible for decreased selective activity (by on average 0.27 ± 0.65 pChEMBL units) on FA10. Among the binding pocket residues, the amino acids (arginine, leucine and tyrosine) at positions 35, 39, 60, 93, 140 and 207 were observed as key contributing residues for selective affinity on these three targets.

Rationalization of activity cliffs of a sulfonamide inhibitor of DNA methyltransferases with induced-fit docking.

Inhibitors of human DNA methyltransferases (DNMT) are of increasing interest to develop novel epi-drugs for the treatment of cancer and other diseases. As the number of compounds with reported DNMT inhibition is increasing, molecular docking is shedding light to elucidate their mechanism of action and further interpret structure-activity relationships. Herein, we present a structure-based rationalization of the activity of SW155246, a distinct sulfonamide compound recently reported as an inhibitor of human DNMT1 obtained from high-throughput screening. We used flexible and induce-fit docking to develop a binding model of SW155246 with a crystallographic structure of human DNMT1. Results were in excellent agreement with experimental information providing a three-dimensional structural interpretation of 'activity cliffs', e.g., analogues of SW155246 with a high structural similarity to the sulfonamide compound, but with no activity in the enzymatic assay.