Biology subjects

Duran-Frigola, M.

Publications and source records attributed to Duran-Frigola, M..

12 recordsLinked to original sources

bioRxiv2024

Integration of diverse bioactivity data into the Chemical Checker compound universe

Chemical signatures encode the physicochemical and structural properties of small molecules into numerical descriptors, forming the basis for chemical comparisons and search algorithms. The increasing availability of bioactivity data has improved compound representations to include biological effects, although bioactivity descriptors are often limited to a few well-documented molecules. To address this issue, we implemented a collection of deep neural networks able to leverage the experimentally determined bioactivity data associated to small molecules and infer the missing bioactivity signatures for any compound of interest. However, unlike static chemical descriptors, these bioactivity signatures dynamically evolve with new data and processing strategies. Here, we present a computational protocol to modify or generate novel bioactivity spaces and signatures, describing the main steps needed to leverage diverse bioactivity data with the current knowledge, as catalogued in the Chemical Checker (CC), using the predefined data curation pipeline. We illustrate the functioning of the protocol through four specific examples, including the incorporation of new compounds to an already existing bioactivity space, a change in the data pre-processing without altering the underlying experimental data, and the creation of two novel bioactivity spaces from scratch, which are completed in under 9 hours using GPU computing. Overall, this protocol offers a guideline for installing, testing and running the CC data integration approach on user-provided data, with the aim of extending the annotation presented for a limited number of small molecules to a larger chemical landscape. Key pointsO_LIThe Chemical Checker is a large collection of processed, harmonized and integrated bioactivity signatures for over 1 million small molecules. Data are organized into 5 levels of increasing biological complexity and curation degrees: from chemical properties to clinical outcomes and from raw data representing explicit knowledge to embedded signatures inferred from observed bioactivity patterns. C_LIO_LIThe Chemical Checker package provides a predefined and versatile framework to integrate user-provided data to the Chemical Checker universe of small molecules and generate novel customized bioactivity signatures. C_LI Key referencesDuran-Frigola et al. "Extending the small-molecule similarity principle to all levels of biology with the Chemical Checker." Nature Biotechnology 38.9 (2020): 1087-1096. Bertoni et al. "Bioactivity descriptors for uncharacterized chemical compounds." Nature Communications 12.1 (2021): 3932.

bioinformatics↗

bioRxiv2024

Learning and actioning general principles of cancer cell drug sensitivity

High-throughput screening platforms for the profiling of drug sensitivity of hundreds of cancer cell lines (CCLs) have generated large datasets that hold the potential to unlock targeted, anti-tumor therapies. In this study, we leveraged these datasets to create predictive models of cancer cells drug sensitivity. To this aim we trained explainable machine learning algorithms by employing cell line transcriptomics to predict the growth inhibitory potential of drugs. We used large language models (LLMs) to expand descriptions of the mechanisms of action (MOA) for each drug starting from available annotations, which were matched to the semantically closest pathways from reference knowledge bases. By leveraging this AI-curated resource, and the interpretability of our model, we demonstrated that pathways enriched for genes crucial for prediction often matched known drug-MOAs and essential genes, suggesting that our models learned the molecular determinants of drug response. Furthermore, we demonstrated that by incorporating only LLM-curated genes associated with MOAs, we enhanced the predictive accuracy of our drug models. To enhance translatability to a clinical setting, we employed a pipeline to align bulk RNAseq from CCLs, used for training the models, to those from patient samples, used for inference. We proved the effectiveness of our approach on TCGA samples, where patients best scoring drugs matched those prescribed for their cancer type. We further showed its usefulness by predicting and experimentally validating effective drugs for the patients of two highly lethal solid tumors, i.e. pancreatic cancer and glioblastoma. In summary, our method facilitates the inference and interpretation of cancer cell line drug sensitivity and holds potential to effectively translate them into new cancer therapeutics. Highlights-Interpretable drug-response prediction models were trained on large scale pharmacogenomics data sets (i.e. GDSC and PRISM). -Large language models were used to enhance the curation of biological pathways associated to drugs MOA -Unbiased interpretation of the models demonstrated learning of drug MOAs and gene essentiality -Inference of TCGA cohort samples recovered mono- and combination cancer drug prescriptions and indicate potential repurposing candidates. -Drug candidates predicted from bulk RNAseq samples of pancreatic cancer and glioblastoma were experimentally validated.

bioinformatics↗

bioRxiv2023

Machine learning approaches identify chemical features for stage-specific antimalarial compounds

Efficacy data from diverse chemical libraries, screened against the various stages of the malaria parasite Plasmodium falciparum, including asexual blood stage (ABS) parasites and transmissible gametocytes, serves as a valuable reservoir of information on the chemical space of compounds that are either active (or not) against the parasite. We postulated that this data can be mined to define chemical features associated with sole ABS activity and/or those that provide additional life cycle activity profiles like gametocytocidal activity. Additionally, this information could provide chemical features associated with inactive compounds, which could eliminate any future unnecessary screening of similar chemical analogues. Therefore, we aimed to use machine learning to identify the chemical space associated with stage-specific antimalarial activity. We collected data from various chemical libraries that were screened against the asexual (126 374 compounds) and sexual (gametocyte) stages of the parasite (93 941 compounds), calculated the compounds molecular fingerprints and trained machine learning models to recognize stage-specific active and inactiv compounds. We were able to build several models that predicts compound activity against ABS and dual-activity against ABS and gametocytes, with Support Vector Machines (SVM) showing superior abilities with high recall (90% and 66%) and low false positive predictions (15% and 1%). This allowed identification of chemical features enriched in active and inactive populations, an important outcome that could be mined for essential chemical features to streamline hit-to-lead optimization strategies of antimalarial candidates. The predictive capabilities of the models held true in diverse chemical spaces, indicating that the ML models are therefore robust and can serve as a prioritization tool to drive and guide phenotypic screening and medicinal chemistry programs. For Table of Contents Graphic Only O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=73 SRC="FIGDIR/small/553339v1_ufig1.gif" ALT="Figure 1"> View larger version (17K): org.highwire.dtl.DTLVardef@6c03d2org.highwire.dtl.DTLVardef@16eeefborg.highwire.dtl.DTLVardef@bcd95org.highwire.dtl.DTLVardef@e60be4_HPS_FORMAT_FIGEXP M_FIG C_FIG

biochemistry↗

bioRxiv2022

First fully-automated AI/ML virtual screening cascade implemented at a drug discovery centre in Africa

We present ZairaChem, an artificial intelligence (AI)- and machine learning (ML)-based tool to train small-molecule activity prediction models. ZairaChem is fully automated, requires low computational resources and works across a broad spectrum of datasets, ranging from whole-cell growth inhibition assays to drug metabolism properties. The tool has been implemented end-to-end at the Holistic Drug Discovery and Development (H3D) Centre, the leading integrated drug discovery unit in Africa, at which no prior AI/ML capabilities were available. We have exploited in-house data collected from over a decade of drug discovery research in malaria and tuberculosis and built models to predict the outcomes of 15 key checkpoint assays. We subsequently deployed these models as a virtual screening cascade at an organisational scale to increase the hit rate of current experimental assays. We show how computational profiling of compounds, prior to synthesis and experimental testing, can increase the rate of progression by up to 40%. Moreover, we demonstrate that the approach can be applied to prioritise small molecules within a chemical series and to assess the likelihood of success of novel chemotypes, promoting efficient usage of limited experimental resources. This project is part of a first-of-its-kind collaboration between the H3D Centre, a research centre operating in a low-resource setting, and the Ersilia Open Source Initiative, a young tech non-profit devoted to building data science capacity in the Global South.

bioinformatics↗

bioRxiv2022

Integrating and formatting biomedical data in the Bioteque, a comprehensive repository of pre-calculated knowledge graph embeddings

Biomedical data is accumulating at a fast pace and integrating it into a unified framework is a major challenge, so that multiple views of a given biological event can be considered simultaneously. Here we present the Bioteque, a resource of unprecedented size and scope that contains pre-calculated biomedical descriptors derived from a gigantic knowledge graph, displaying more than 450 thousand biological entities and 30 million relationships between them. The Bioteque integrates, harmonizes, and formats data collected from over 150 data sources, including 12 biological entities (e.g., genes, diseases, drugs) linked by 67 types of associations (e.g., drug treats disease, gene interacts with gene). We show how Bioteque descriptors facilitate the assessment of high-throughput protein-protein interactome data, the prediction of drug response and new repurposing opportunities, and demonstrate that they can be used off-the-shelf in downstream machine learning tasks without loss of performance with respect to using original data. The Bioteque thus offers a thoroughly processed, tractable, and highly optimized assembly of the biomedical knowledge available in the public domain.

bioinformatics↗

bioRxiv2021

Identification and drug-induced reversion of molecular signatures of Alzheimers disease onset and progression in AppNL-G-F, AppNL-F and 3xTg-AD mouse models

Alzheimers disease (AD) is the most common form of dementia. Over fifty years of intense research have revealed many key elements of the biology of this neurodegenerative disorder. However, our understanding of the molecular bases of the disease is still incomplete, and the medical treatments available for AD are mainly symptomatic and hardly effective. Indeed, the robustness of biological systems has revealed that the modulation of a single target is unlikely to yield the desired outcome and we should therefore move from gene-centric to systemic therapeutic strategies. Here we present the complete characterization of three murine models of AD at different stages of the disease (i.e. onset, progression and advanced). To identify genotype-to-phenotype relationships, we combine the cognitive assessment of these mice with histological analyses and full transcriptional and protein quantification profiling of the hippocampus. Comparison of the gene and protein expression trends observed in AD progression and physiological aging revealed certain commonalities, such as the upregulation of microglial and inflammation markers. However, although AD models show accelerated aging, other factors specifically associated with A{beta} pathology are involved. Despite the clear correlation between mRNA and protein levels of the dysregulated genes, we discovered a few proteins whose abundance increases with AD progression, while the corresponding transcript levels remain stable. Indeed, we show that at least two of these proteins, namely lfit3 and Syt11, co-localize with A{beta} plaques in the brain. Finally, we derived specific A{beta}-related molecular AD signatures and looked for drugs able to globally revert them. We found two NSAIDs (dexketoprofen and etodolac) and two anti-hypertensives (penbutolol and bendroflumethiazide) that overturn the cognitive impairment in AD mice while reducing A{beta} plaques in the hippocampus and partially restoring the physiological levels of AD signature genes to wild-type levels. TeaserThe comprehensive characterization of three AD mouse models reveals disease signatures that we used to identify approved drugs able to modify the etiology of the pathology and overturn cognitive impairment.

neuroscience↗

bioRxiv2020

A Community Challenge for Pancancer Drug Mechanism of Action Inference from Perturbational Profile Data

The Columbia Cancer Target Discovery and Development (CTD2) Center has developed PANACEA (PANcancer Analysis of Chemical Entity Activity), a collection of dose-response curves and perturbational profiles for 400 clinical oncology drugs in cell lines selected to optimally represent 19 cancer subtypes. This resource, developed to study tumor-specific drug mechanism of action, was instrumental in hosting a DREAM Challenge to assess computational models for de novo drug polypharmacology prediction. Dose-response and perturbational profiles for 32 kinase inhibitors were provided to 21 participating teams, who did not know the identity or nature of the compounds, and they were asked to predict high-affinity binding among ~1,300 possible protein targets. Best performing methods leveraged both gene expression profile similarity analysis, and deep-learning methodologies trained on individual datasets. This study lays the foundation for future integrative analyses of pharmacogenomic data, reconciliation of polypharmacology effects in different tumor contexts, and insights into network-based assessment of context-specific drug mechanism of action.

cancer biology↗

bioRxiv2020

Bioactivity descriptors for uncharacterized compounds

Chemical descriptors encode the physicochemical and structural properties of small molecules, and they are at the core of chemoinformatics. The broad release of bioactivity data has prompted enriched representations of compounds, reaching beyond chemical structures and capturing their known biological properties. Unfortunately, bioactivity descriptors are not available for most small molecules, which limits their applicability to a few thousand well characterized compounds. Here we present a collection of deep neural networks able to infer bioactivity signatures for any compound of interest, even when little or no experimental information is available for them. Our signaturizers relate to bioactivities of 25 different types (including target profiles, cellular response and clinical outcomes) and can be used as drop-in replacements for chemical descriptors in day-to-day chemoinformatics tasks. Indeed, we illustrate how inferred bioactivity signatures are useful to navigate the chemical space in a biologically relevant manner, unveiling higher-order organization in natural product collections, and to enrich mostly uncharacterized chemical libraries for activity against the drug-orphan target Snail1. Moreover, we implement a battery of signature-activity relationship (SigAR) models and show a substantial improvement in performance, with respect to chemistry-based classifiers, across a series of biophysics and physiology activity prediction benchmarks.

bioinformatics↗

bioRxiv2019

Personalized Cancer Therapy Prioritization Based on Driver Alteration Co-occurrence Patterns

Identification of actionable genomic vulnerabilities is the cornerstone of precision oncology. Based on a large-scale drug screening in patient derived-xenografts, we uncover connections between driver gene alterations, derive Driver Co-Occurrence (DCO) networks, and relate these to drug sensitivity. Our collection of 53 drug response predictors attained an average balanced accuracy of 58% in a cross-validation setting, which rose to a 66% for the subset of high-confidence predictions. Morevover, we experimentally validated 12 out of 14 de novo predictions in mice. Finally, we adapted our strategy to obtain drug-response models from patients progression free survival data. By revealing unexpected links between oncogenic alterations, our strategy can increase the clinical impact of genomic profiling.

cancer biology↗

bioRxiv2019

Extending the small molecule similarity principle to all levels of biology

We present the Chemical Checker (CC), a resource that provides processed, harmonized and integrated bioactivity data on 800,000 small molecules. The CC divides data into five levels of increasing complexity, ranging from the chemical properties of compounds to their clinical outcomes. In between, it considers targets, off-targets, perturbed biological networks and several cell-based assays such as gene expression, growth inhibition and morphological profilings. In the CC, bioactivity data are expressed in a vector format, which naturally extends the notion of chemical similarity between compounds to similarities between bioactivity signatures of different kinds. We show how CC signatures can boost the performance of drug discovery tasks that typically capitalize on chemical descriptors, including target identification and library characterization. Moreover, we demonstrate and experimentally validate that CC signatures can be used to reverse and mimic biological signatures of disease models and genetic perturbations, options that are otherwise impossible using chemical information alone.

bioinformatics↗

bioRxiv2019

A reference map of the human protein interactome

Global insights into cellular organization and function require comprehensive understanding of interactome networks. Similar to how a reference genome sequence revolutionized human genetics, a reference map of the human interactome network is critical to fully understand genotype-phenotype relationships. Here we present the first human \"all-by-all\" binary reference interactome map, or \"HuRI\". With ~53,000 high-quality protein-protein interactions (PPIs), HuRI is approximately four times larger than the information curated from small-scale studies available in the literature. Integrating HuRI with genome, transcriptome and proteome data enables the study of cellular function within essentially any physiological or pathological cellular context. We demonstrate the use of HuRI in identifying specific subcellular roles of PPIs and protein function modulation via splicing during brain development. Inferred tissue-specific networks reveal general principles for the formation of cellular context-specific functions and elucidate potential molecular mechanisms underlying tissue-specific phenotypes of Mendelian diseases. HuRI thus represents an unprecedented, systematic reference linking genomic variation to phenotypic outcomes.

systems biology↗

bioRxiv2018

Encircling the regions of the pharmacogenomic landscape that determine drug response

The integration of large-scale drug sensitivity screens and genome-wide experiments is changing the field of pharmacogenomics, revealing molecular determinants of drug response without the need for a priori, hypothesis-driven assumptions about drug action. In particular, transcriptomic signatures of drug sensitivity may guide drug repositioning, the discovery of synergistic drug combinations and suggest new therapeutic biomarkers. However, the inherent complexity of transcriptomic signatures, with thousands of genes differentially expressed, makes them hard to interpret, giving poor mechanistic insights and hampering translation to the clinics. Here we show how network biology can help simplify transcriptomic drug signatures, filtering out irrelevant genes, accounting for tissue-specific biases and ultimately yielding functionally-coherent, less noisy drug modules. We successfully analyzed 170 drugs tested in 637 cancer cell lines, proving a broad applicability of our approach and evincing an intimate relationship between modules gene expression levels and drugs mechanisms of action. Further, we have characterized multiple aspects of our transcriptomic modules. As a result, the drugs included in this study are now annotated well beyond the reductionist (target-centered) view.\n\nAuthor SummaryLarge scale pharmacogenomics studies performed with hundreds of cell lines offer a means to link the molecular features of the cells to their response to drug treatments. Unfortunately, simple drug-gene correlations are usually not enough to consistently identify what gene expression patterns will determine drug sensitivity, as the tissue of origin of the cells, together with the expression of e.g. membrane transporters can greatly confound the analysis. To ameliorate these biases, we have devised a network-based strategy that selects genes that are both well correlated to drug response and closeby in the human protein interaction network. Reassuringly, we have confirmed that our identified drug sensitivity modules are tightly connected to the mechanisms of action of the drugs. Moreover, while our modules have no more than 100 genes, they retain the predictive power of the much larger gene signatures that are typically obtained by drug-gene correlations alone. Here, we release the characteristic modules for almost 200 drugs, in a format that is suitable for downstream bioinformatics analyses such as gene-set enrichment analysis.

bioinformatics↗