Matching single cells across modalities with contrastive learning and optimal transport.

Understanding the interactions between the biomolecules that govern cellular behaviors remains an emergent question in biology. Recent advances in single-cell technologies have enabled the simultaneous quantification of multiple biomolecules in the same cell, opening new avenues for understanding cellular complexity and heterogeneity. Still, the resulting multimodal single-cell datasets present unique challenges arising from the high dimensionality and multiple sources of acquisition noise. Computational methods able to match cells across different modalities offer an appealing alternative towards this goal. In this work, we propose MatchCLOT, a novel method for modality matching inspired by recent promising developments in contrastive learning and optimal transport. MatchCLOT uses contrastive learning to learn a common representation between two modalities and applies entropic optimal transport as an approximate maximum weight bipartite matching algorithm. Our model obtains state-of-the-art performance on two curated benchmarking datasets and an independent test dataset, improving the top scoring method by 26.1% while preserving the underlying biological structure of the multimodal data. Importantly, MatchCLOT offers high gains in computational time and memory that, in contrast to existing methods, allows it to scale well with the number of cells. As single-cell datasets become increasingly large, MatchCLOT offers an accurate and efficient solution to the problem of modality matching.

ATHENA: analysis of tumor heterogeneity from spatial omics measurements.

SUMMARY: Tumor heterogeneity has emerged as a fundamental property of most human cancers, with broad implications for diagnosis and treatment. Recently, spatial omics have enabled spatial tumor profiling, however computational resources that exploit the measurements to quantify tumor heterogeneity in a spatially aware manner are largely missing. We present ATHENA (Analysis of Tumor HEterogeNeity from spAtial omics measurements), a computational framework that facilitates the visualization, processing and analysis of tumor heterogeneity from spatial omics measurements. ATHENA uses graph representations of tumors and bundles together a large collection of established and novel heterogeneity scores that quantify different aspects of the complexity of tumor ecosystems. AVAILABILITY AND IMPLEMENTATION: ATHENA is available as a Python package under an open-source license at: https://github.com/AI4SCR/ATHENA. Detailed documentation and step-by-step tutorials with example datasets are also available at: https://ai4scr.github.io/ATHENA/. The data presented in this article are publicly available on Figshare at https://figshare.com/articles/dataset/zurich_pkl/19617642/2. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

VenomPred: A Machine Learning Based Platform for Molecular Toxicity Predictions.

The use of in silico toxicity prediction methods plays an important role in the selection of lead compounds and in ADMET studies since in vitro and in vivo methods are often limited by ethics, time, budget and other resources. In this context, we present our new web tool VenomPred, a user-friendly platform for evaluating the potential mutagenic, hepatotoxic, carcinogenic and estrogenic effects of small molecules. VenomPred platform employs several in-house Machine Learning (ML) models developed with datasets derived from VEGA QSAR, a software that includes a comprehensive collection of different toxicity models and has been used as a reference for building and evaluating our ML models. The results showed that our models achieved equal or better performance than those obtained with the reference models included in VEGA QSAR. In order to improve the predictive performance of our platform, we adopted a consensus approach combining the results of different ML models, which was able to predict chemical toxicity better than the single models. This improved method was thus implemented in the VenomPred platform, a freely accessible webserver that takes the SMILES (Simplified Molecular-Input Line-Entry System) strings of the compounds as input and sends the prediction results providing a probability score about their potential toxicity.

Comparative evaluation of DNA-encoded chemical selections performed using DNA in single-stranded or double-stranded format.

DNA-encoded chemical libraries (DEL) are increasingly being used for the discovery and optimization of small organic ligands to proteins of biological or pharmaceutical interest. The DNA fragments, that serve as amplifiable identification barcodes for individual compounds in the library, are typically used in double-stranded DNA format. To the best of our knowledge, a direct comparison of DEL selections featuring DNA in either single- or double-stranded DNA format has not yet been reported. In this article, we describe a comparative evaluation of selections with two DEL libraries (named GB-DEL and NF-DEL), based on different chemical designs and produced in both single- and double-stranded DNA format. The libraries were selected in identical conditions against multiple protein targets, revealing comparable and reproducible fingerprints for both types of DNA formats. Surprisingly, selections performed with single-stranded DNA barcodes exhibited improved enrichment factors compared to double-stranded DNA. Using high-affinity ligands to carbonic anhydrase IX as benchmarks for selection performance, we observed an improved selectivity for the NF-DEL library (on average 2-fold higher enrichment factors) in favor of single-stranded DNA. The enrichment factors were even higher for the GB-DEL selections (approximately 5-fold), compared to the same library in double-stranded DNA format. Collectively, these results indicate that DEL libraries can conveniently be synthesized and screened in both single- and double-stranded DNA format, but single-stranded DNA barcodes typically yield enhanced enrichment factors.

Development of a cheminformatics platform for selectivity analyses of carbonic anhydrase inhibitors.

The selectivity for a specific human Carbonic Anhydrase (hCA) isoform is an important property a hCA inhibitor (CAI) should be endowed with, in order to constitute a valuable therapeutic tool for the treatment of a desired pathology. In this context, we developed a chemoinformatic platform that allows the analysis of the structure and selectivity profile of known CAIs reported in literature, with the aim of identifying structural motifs connected to ligand selectivity, thus providing useful guidelines for the design of novel ligands selective for the desired hCA isoform. The platform is able to perform ultrafast structure and selectivity analyses through ligand fingerprint similarity, with no need of structural information about the target receptor and ligands' binding mode. It is easily accessible to the non-expert user through the implementation of a KNIME Analytic Platform workflow and could be extended to analyze the selectivity profile of known ligands of different target proteins.

Extensive Reliability Evaluation of Docking-Based Target-Fishing Strategies.

The development of target-fishing approaches, aimed at identifying the possible protein targets of a small molecule, represents a hot topic in medicinal chemistry. A successful target-fishing approach would allow for the elucidation of the mechanism of action of all therapeutically interesting compounds for which the actual target is still unknown. Moreover, target-fishing would be essential for preventing adverse effects of drug candidates, by predicting their potential off-targets, and it would speed up drug repurposing campaigns. However, due to the huge number of possible protein targets that a small-molecule might interact with, experimental target-fishing approaches are out of reach. In silico target-fishing represents a valuable alternative, and examples of receptor-based approaches, exploiting the large number of crystallographic protein structures determined to date, have been reported in the literature. To the best of our knowledge, no proper evaluation of such approaches is, however, reported yet. In the present work, we extensively assessed the reliability of docking-based target-fishing strategies. For this purpose, a set of X-ray structures belonging to different targets was selected, and a dataset of compounds, including 10 experimentally active ligands for each target, was created. A target-fishing benchmark database was then obtained, and used to assess the performance of 13 different docking procedures, in identifying the correct target of the dataset ligands. Moreover, a consensus docking-based target-fishing strategy was developed and evaluated. The analysis highlighted that specific features of the target proteins could affect the reliability of the protocol, which however, proved to represent a valuable tool in the proper applicability domain. Our study represents the first extensive performance assessment of docking-based target-fishing approaches, paving the way for the development of novel efficient receptor-based target fishing strategies.

Development of a Fingerprint-Based Scoring Function for the Prediction of the Binding Mode of Carbonic Anhydrase II Inhibitors.

Carbonic anhydrase II (CAII) is a zinc-containing metalloenzyme whose aberrant activity is associated with various diseases such as glaucoma, osteoporosis, and different types of tumors; therefore, the development of CAII inhibitors, which can represent promising therapeutic agents for the treatment of these pathologies, is a current topic in medicinal chemistry. Molecular docking is a commonly used tool in structure-based drug design of enzyme inhibitors. However, there is still a need for improving docking reliability, especially in terms of scoring functions, since the complex pattern of energetic contributions driving ligand⁻protein binding cannot be properly described by mathematical functions only including approximated energetic terms. Here we report a novel CAII-specific fingerprint-based (IFP) scoring function developed according to the ligand⁻protein interactions detected in the CAII-inhibitor co-crystal structures of the most potent CAII ligands. Our IFP scoring function outperformed the ability of Autodock4 scoring function to identify native-like docking poses of CAII inhibitors and thus allowed a considerable improvement of docking reliability. Moreover, the ligand⁻protein interaction fingerprints showed a useful application in the binding mode analysis of structurally diverse CAII ligands.

Development of terphenyl-2-methyloxazol-5(4H)-one derivatives as selective reversible MAGL inhibitors.

Monoacylglycerol lipase is a serine hydrolase that plays a major role in the degradation of the endocannabinoid neurotransmitter 2-arachidonoylglycerol. A wide number of MAGL inhibitors are reported in literature; however, many of them are characterised by an irreversible mechanism of action and this behavior determines an unwanted chronic MAGL inactivation, which acquires a functional antagonism of the endocannabinoid system. The possible use of reversible MAGL inhibitors has only recently been explored, due to the lack of known compounds possessing efficient reversible inhibitory activities. In this work, we report a new series of terphenyl-2-methyloxazol-5(4H)-one derivatives characterised by a reversible MAGL-inhibition mechanism. Among them, compound 20b showed to be a potent MAGL reversible inhibitor (IC50 = 348 nM) with a good MAGL/FAAH selectivity. Furthermore, this compound showed antiproliferative activities against two different cancer cell lines that overexpress MAGL.

Reliability analysis and optimization of the consensus docking approach for the development of virtual screening studies.

Ligand-protein docking is one of the most common techniques used in virtual screening campaigns. Despite the large number of docking software available, there is still the need of improving the efficacy of docking-based virtual screenings. To date, only very few studies evaluated the possibility of combining the results of different docking methods to achieve higher success rates in virtual screening studies (consensus docking). In order to better understand the range of applicability of this approach, we carried out an extensive enriched database analysis using the DUD dataset. The consensus docking protocol was then refined by applying modifications concerning the calculation of pose consensus and the combination of docking methods included in the procedure. The results obtained suggest that this approach performs as well as the best available methods found in literature, confirming the idea that this procedure can be profitably used for the identification of new hit compounds.

Identification of a new STAT3 dimerization inhibitor through a pharmacophore-based virtual screening approach.

Signal transducer and activator of transcription 3 (STAT3) plays an essential role in cell growth regulation and survival. An aberrant STAT3 activation and/or expression is implied in various solid and blood tumors as well as in other pathologies like rheumatoid arthritis and pulmonary fibrosis, thus making the search for STAT3 inhibitors a growing field of study. With the aim of identifying new inhibitors of STAT3 dimerization, we screened a database including more than 1 320 000 commercially available compounds using a receptor-based pharmacophore model comprising the key protein-protein interactions identified in the STAT3 dimer and refining the search through docking and molecular dynamic simulations studies. STAT3 binding assays revealed a significant STAT3 inhibitory activity and selectivity versus Grb2 for one of the four top-scored compounds, thus verifying the reliability of the virtual screening workflow. Moreover, such compound could already be considered as a lead for the development of new and more potent STAT3 dimerization inhibitors.

4-Aryliden-2-methyloxazol-5(4H)-one as a new scaffold for selective reversible MAGL inhibitors.

This study reports on a preliminary structure-activity relationship exploration of 4-aryliden-2-methyloxazol-5(4H)-one-based compounds as MAGL/FAAH inhibitors. Our results highlight that this scaffold may serve for the development of selective MAGL inhibitors. A 69-fold selectivity against MAGL over FAAH was achieved for compound 16b (MAGL and FAAH IC(50) = 1.6 and 111 µM, respectively). Furthermore, the best compound behaved as a reversible ligand and showed promising antiproliferative activity in cancer cells.

Application of a FLAP-consensus docking mixed strategy for the identification of new fatty acid amide hydrolase inhibitors.

Fatty acid amide hydrolase (FAAH) is the principal responsible for the termination of anandamide signaling, a major actor of the endocannabinoid system. The indirect stimulation of endocannabinoid responses achieved through FAAH inhibition can represent a valid pharmacological strategy for the treatment of neurodegenerative and neuroinflammatory diseases such as multiple sclerosis, Alzheimer's, Huntington's, and Parkinson's diseases, as well as rheumatoid arthritis, gastrointestinal inflammatory states, anxiety, and other pathologies. With the aim of identifying new noncovalent FAAH inhibitors and also experimentally validating the reliability of the recently reported consensus docking approach, we filtered a commercial database of about 1 million compounds by using a mixed FLAP (fingerprints for ligands and proteins) consensus docking approach. Enzymatic assays showed FAAH inhibitory activity and selectivity versus MAGL for 8 out of the 10 top ranked compounds, with IC50 values in the low micromolar range for the two most active compounds. These results demonstrate the reliability of the virtual screening strategy and constitute an experimental validation of the consensus docking approach. Moreover, the two most active compounds described could represent promising leads for the development of high potent noncovalent FAAH inhibitors.

Receptor-based virtual screening evaluation for the identification of estrogen receptor ß ligands.

In this paper, a receptor-based virtual screening study for the identification of estrogen receptor ß (ERß) ligands was developed. Starting from a commercial database of 400,000 molecules, only six compounds resulted to be potential active ligands of ERß. Interestingly, all the six molecules possess scaffolds that had already been reported in known ERß ligands. Therefore, the results obtained herein confirm the reliability of our virtual screening procedure, thus encouraging the application of this protocol to larger commercial databases in order to identify new ERß ligands.

Development and validation of a docking-based virtual screening platform for the identification of new lactate dehydrogenase inhibitors.

The human muscle isoform of lactate dehydrogenase (hLDH5) is one of the key enzymes of the glycolytic process. It is overexpressed in metastatic cancer cells and is linked to the vitality of tumors in hypoxic conditions. With the aim of identifying new hLDH5 inhibitors, a fully automated docking-based virtual screening platform was developed by considering different protein conformations and the consensus docking strategy. In order to verify the reliability of the reported platform, a small database of about 10,000 compounds was filtered by using this method, and the top-ranked compounds were tested for their hLDH5 inhibition activity. Enzymatic assays revealed that, among the ten selected compounds, two proved to efficiently inhibit enzyme activity with IC50 values in the micromolar range. These results demonstrate the validity of the methodologies we followed, encouraging the application of larger virtual screening studies and further refinements of the platform. Furthermore, the two active compounds herein described may be considered as interesting leads for the development of new and more efficient LDH inhibitors.

Extensive consensus docking evaluation for ligand pose prediction and virtual screening studies.

Molecular docking strategies are one of the most widely used techniques for predicting the binding mode of a ligand and for obtaining new hits in virtual screening studies. In order to improve the accuracy of this approach, we tested the reliability of applying a consensus docking protocol by combining ten different docking procedures. The analysis was carried out in terms of consensus cross-docking and by using an enriched database. The results highlight that from a qualitative point of view consensus docking is able to predict the ligand binding pose better than the single docking programs and is also able to give hints concerning the reliability of the docking pose. With regard to the virtual screening studies, consensus docking was evaluated for three different targets of the Directory of Useful Decoys (DUD), and the obtained results suggest that this approach performs as well as the best available methods found in the literature, therefore supporting the idea that this procedure can be profitably applied for the identification of new hits.

Identification and characterization of a new reversible MAGL inhibitor.

Monoacylglycerol lipase is a serine hydrolase that play a major role in the degradation of 2-arachidonoylglycerol, an endocannabinoid neurotransmitter implicated in several physiological processes. Recent studies have shown the possible role of MAGL inhibitors as anti-inflammatory, anti-nociceptive and anti-cancer agents. The use of irreversible MAGL inhibitors determined an unwanted chronic MAGL inactivation, which acquires a functional antagonism function of the endocannabinoid system. However, the application of reversible MAGL inhibitors has not yet been explored, mainly due to the scarcity of known compounds possessing efficient reversible inhibitory activities. In this study we reported the first virtual screening analysis for the identification of reversible MAGL inhibitors. Among the screened compounds, the (4-(4-chlorobenzoyl)piperidin-1-yl)(4-methoxyphenyl)methanone (CL6a) is a promising reversible MAGL inhibitor lead (Ki=8.6µM), which may be used for the future development of a new class of MAGL inhibitors. Furthermore, the results demonstrate the validity of the methodologies that we followed, encouraging additional screenings of other commercial databases.

Highly pure DNA-encoded chemical libraries by dual-linker solid-phase synthesis.

The first drugs discovered using DNA-encoded chemical library (DEL) screens have entered late-stage clinical development. However, DEL technology as a whole still suffers from poor chemical purity resulting in suboptimal performance. In this work, we report a technique to overcome this issue through self-purifying release of the DEL after magnetic bead-based synthesis. Both the first and last building blocks of each assembled library member were linked to the beads by tethers that could be cleaved by mutually orthogonal chemistry. Sequential cleavage of the first and last tether, with washing in between, ensured that the final library comprises only the fully complete compounds. The outstanding purity attained by this approach enables a direct correlation of chemical display and encoding, allows for an increased chemical reaction scope, and facilitates the use of more diversity elements while achieving greatly improved signal-to-noise ratios in selections.

Dual targeting of the Warburg effect with a glucose-conjugated lactate dehydrogenase inhibitor.

Effective glucose diet: We report the development and activity of glucose-conjugated LDH-A inhibitors designed for dual targeting of the Warburg effect (elevated glucose uptake and glycolysis) in cancer cells. Glycoconjugation could be applied to inhibitors of many enzymes involved in glycolysis or tumor metabolism.

Oxime-based inhibitors of glucose transporter 1 displaying antiproliferative effects in cancer cells.

An analysis of the main pharmacophoric features present in the still limited number of inhibitors of glucose transporter GLUT1 led to the identification of new oxime-based inhibitors, which proved to be able to efficiently hinder glucose uptake and cell growth in H1299 lung cancer cells. The most important interactions of a representative inhibitor were indicated by a novel computational model of GLUT1, which was purposely developed to explain these results and to provide useful indications for the design and the development of new and more efficient GLUT1 inhibitors.

Assessing the differential action on cancer cells of LDH-A inhibitors based on the N-hydroxyindole-2-carboxylate (NHI) and malonic (Mal) scaffolds.

A head-to-head study of representative examples of N-hydroxyindole-2-carboxylates (NHI) and malonic derivatives (Mal) as LDH-A inhibitors was conducted, comparing the enzyme inhibition potency, cellular uptake, reduction of lactate production in cancer cells and anti-proliferative activity. Among the compounds tested, methyl 1-hydroxy-6-phenyl-4-(trifluoromethyl)-1H-indole-2-carboxylate (2, NHI-2), a methyl ester belonging to the NHI class, displayed optimal properties in the cell-based assays, proving to be an efficient anti-glycolytic agent against cancer cells.