Biology subjects

Aloy, P.

Publications and source records attributed to Aloy, P..

At least 19 recordsLinked to original sources

bioRxiv2026

Task-adapted biological foundation models uncover perturbation-centric representations

Foundation models have emerged as powerful tools for learning transferable representations of biological systems, yet their latent spaces are typically optimized to capture cellular state rather than the effects of perturbations. Here, we demonstrate that a biological foundation model can be repurposed to learn a fundamentally different representation by changing its learning objective. We fine-tuned scGPT, a transformer pre-trained on over 30 million single-cell transcriptomes, on more than three million LINCS L1000 perturbation profiles using a supervised objective that predicts perturbation identity. This transformed the latent space into a perturbation-centric representation that aligned transcriptional responses induced by the same chemical or genetic perturbation across heterogeneous experimental conditions. Fine-tuned embeddings substantially outperformed both gene expression profiles and the original pre-trained model, recovering 85-100% of perturbations within the top 100 nearest neighbors and increasing perturbation classification accuracy from 10-19% to 25-49%. Remarkably, although the model was trained exclusively to recognize perturbation identity, the learned representation spontaneously captured orthogonal biological relationships never provided during training, including chemical similarity (AUROC up to 0.81), mechanisms of action (Hit@10 up to 100%), compound-target relationships (AUROC up to 0.74), and functional relationships between genetic perturbations. The resulting embedding space enabled mechanism-of-action annotation of nearly 12,000 previously uncharacterized compounds, prioritization of target-related chemical-genetic associations, and contextualization of unseen perturbations and external transcriptomic datasets. Together, our results establish objective-driven adaptation as a general strategy for repurposing biological foundation models to learn reusable representations of complex biological phenomena.

bioinformatics↗

bioRxiv2025

Phenotypic AI-based design of cell-specific small molecule cytotoxics

Drug discovery is being transformed by artificial intelligence, which enables the exploration of vast chemical spaces and the generation of novel compounds with tailored properties. While most generative approaches focus on optimizing physicochemical features or protein-binding affinities, their application to phenotypic drug discovery remains limited. Here we present an integrated framework to design small molecules with selective cytotoxic effects in pancreatic cancer cells that combines high-throughput cytotoxicity screening, bioactivity signature-based machine learning predictors, and reinforcement learning-driven generative models. We screened over 11,000 compounds across six pancreatic cancer and two control cell lines, confirming 392 hits, and used these data to train cell line-specific classifiers that outperformed fingerprint-based models. We embedded these predictors into the REINVENT platform to generate molecules with desired cytotoxicity profiles, including challenging cases of distinguishing between pancreatic cancer lines with highly similar molecular backgrounds. We generated between 37 and 137 high- scoring candidates in seven different cell selectivity exercises, which were structurally distant from the active compounds in the training set. We tested 45 close analogues to the AI-designed molecules, and 20 of them showed the desired cell-specific cytotoxic effects. Indeed, we validated at least one compound in 5 of the 7 exercises tested, showing over a large increase in hit discovery rate compared to the HTS. Additionally, we benchmarked our strategy against direct predictions on a library of over 150,000 compounds, skipping the generative step, finding that the fraction of hits, and more importantly the dose-response validation, is markedly higher in the AI- designed molecules. Overall, our study demonstrates the feasibility of combining predictive and generative AI methods to design molecules with complex phenotypic outcomes in a target- agnostic manner, paving the way for next-generation phenotype-driven drug discovery strategies. O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=68 SRC="FIGDIR/small/682546v1_ufig1.gif" ALT="Figure 1"> View larger version (18K): org.highwire.dtl.DTLVardef@119daf2org.highwire.dtl.DTLVardef@7579b9org.highwire.dtl.DTLVardef@8a62f8org.highwire.dtl.DTLVardef@167a6d5_HPS_FORMAT_FIGEXP M_FIG C_FIG

bioinformatics↗

bioRxiv2025

The modifiers that cause changes in gene essentiality

The phenotype of a mutation often differs across genetically distinct individuals. In the most extreme case, a gene can be essential for viability in one genetic background, but dispensable in another. Although genetic context-dependency of mutant phenotypes is frequently observed, the underlying causes often remain elusive. Here, we investigated the genetic changes responsible for differences in gene essentiality across 18 genetically diverse natural yeast strains. First, we identified 39 genes that were essential in the laboratory reference strain but not required for viability in at least one other genetic background, suggesting that the natural strain contained suppressor variants that could bypass the need for the essential gene. We then mapped and validated the causal bypass suppressor variants using bulk segregant analysis and allele replacements. Bypass suppression was generally driven by a single modifier gene that tended to differ between genetic backgrounds. The suppressors often indirectly counteracted the effect of deleting the essential gene, for instance by changing the transcriptome of a cell. Context-dependent essential genes and their bypass suppressors were frequently co-mutated across 1,011 yeast isolates and identified naturally occurring evolutionary trajectories. Overall, our results highlight the relatively high frequency of bypass suppression in natural populations, as well as the underlying variants and mechanisms. A thorough understanding of the causes of genetic background effects is crucial for the interpretation of genotype-to-phenotype relationships, including those associated with human disease.

genomics↗

bioRxiv2025

Systematic assessment of microenvironment-dependent transcriptional patterns and intercellular communication

Understanding cell-cell communication and its dependence on spatial organization is critical for unraveling tissue complexity and organ function. This study integrates single-cell RNA sequencing (scRNA-seq) with spatial transcriptomics (ST) to systematically assess how spatial niches influence gene expression and intercellular communication. Using breast cancer, brain cortex, and heart datasets, our analyses reveal limited global transcriptional changes in cells depending on their spatial microenvironment, with differential gene expression observed in less than half the samples explored. Moreover, cell-cell communication predictions, derived from ligand-receptor pairs, exhibit minimal correlation with spatial colocalization of cell types. Overall, our study underscores the limitations of using scRNA-seq data to capture niche-specific molecular interactions, even when spatial information is leveraged, and it highlights the need for novel strategies to refine our understanding of intercellular communication dynamics at molecular level.

bioinformatics↗

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

Inhibitor-induced supercharging of kinase turnover via endogenous proteolytic circuits

Targeted protein degradation has emerged as a promising new pharmacological strategy. Traditionally, it relies on small molecules that induce proximity between a target protein and an E3 ubiquitin ligase to prompt target ubiquitination and degradation by the proteasome. Sporadic reports indicated that ligands designed to inhibit a target can also induce its destabilization. Among others, this has repeatedly been observed for kinase inhibitors. However, we lack an understanding of the frequency, generalizability, and mechanistic underpinnings of these phenomena. To address this knowledge gap, we generated dynamic abundance profiles of 98 kinases after cellular perturbations with 1570 kinase inhibitors, revealing 160 selective instances of inhibitor-induced kinase destabilization. Kinases prone to degradation are frequently annotated as HSP90 clients, thus affirming chaperone deprivation as an important route of destabilization. However, detailed investigation of inhibitor-induced degradation of LYN, BLK and RIPK2 revealed a differentiated, common mechanistic logic where inhibitors function by inducing a kinase state that is more efficiently cleared by endogenous degradation mechanisms. Mechanistically, effects can manifest by ligand-induced changes in cellular activity, localization, or multimerization which may be triggered by direct target engagement or network effects. Collectively, our data suggest that inhibitor-induced kinase degradation is a common event and positions supercharging of endogenous degradation circuits as an alternative to classical proximity-inducing degraders.

molecular biology↗

bioRxiv2024

Systematic study of hybrid triplex topology and stability suggests a general triplex-mediated regulatory mechanism

By combining in-silico, biophysical and in-cellulo experiments, we decipher the topology, physical and potential biological properties of hybrid-parallel nucleic acids triplexes; an elusive structure at the basis of life. We found that hybrid triplex topology follows a stability order: r(Py)-d(Pu){middle dot}r(Py)> r(Py)-d(Pu){middle dot}d(Py)> d(Py)-d(Pu){middle dot}d(Py)> d(Py)-d(Pu){middle dot}r(Py). The r(Py)-d(Pu){middle dot}d(Py) triplex is expected to be the preferred in the cell as it avoids the need to open the duplex reducing the torsional stress required for triplex formation in the r(Py)-d(Pu){middle dot}r(Py) topology. Upon a massive collection of melting data, we have created the first predictor for hybrid triplex stability. Leveraging this predictor, we conducted a comprehensive scan to assess the likelihood of the human genome and transcriptome to engage in triplex formation. Our findings unveil a remarkable inclination - of both the human genome and transcriptome - to generate hybrid triplex formation, particularly within untranslated (UTRs) and regulatory regions, thereby corroborating the existence of a triplex-mediated regulatory mechanism. Furthermore, we found a correlation between nucleosome linkers and TFS which agree with a putative role of triplexes in arranging chromatin structure and local/global level.

molecular biology↗

bioRxiv2024

Comprehensive detection and characterization of human druggable pockets through novel binding site descriptors

Druggable pockets are protein regions that have the ability to bind organic small molecules, and their characterization is essential in target-based drug discovery. However, strategies to derive pocket descriptors are scarce and usually exhibit limited applicability. Here, we present PocketVec, a novel approach to generate pocket descriptors for any protein binding site of interest through the inverse virtual screening of lead-like molecules. We assess the performance of our descriptors in a variety of scenarios, showing that it is on par with the best available methodologies, while overcoming some important limitations. In parallel, we systematically search for druggable pockets in the folded human proteome, using experimentally determined protein structures and AlphaFold2 models, identifying over 32,000 binding sites in more than 20,000 protein domains. Finally, we derive PocketVec descriptors for each small molecule binding site and run an all-against-all similarity search, exploring over 1.2 billion pairwise comparisons. We show how PocketVec descriptors facilitate the identification of druggable pocket similarities not revealed by structure- or sequence-based comparisons. Indeed, our analyses unveil dense clusters of similar pockets in distinct proteins for which no inhibitor has yet been crystalized, opening the door to strategies to prioritize the development of chemical probes to cover the druggable space.

bioinformatics↗

bioRxiv2024

Stereochemically-aware bioactivity descriptors for uncharacterized chemical compounds

We recently presented a set of deep neural networks to generate bioactivity descriptors associated to small molecules (i.e. Signaturizers), capturing their effects at increasing levels of biological complexity (i.e. from protein targets to clinical outcomes)1. However, such models were trained on 2D representations of molecules and are thus unable to capture key differences in the activity of stereoisomers. Now, we systematically assess the relationship between stereoisomerism and bioactivity on over 1M compounds, finding that a very significant fraction ([~]40%) of spatial isomer pairs show, to some extent, distinct bioactivities. We then used these data to train a second generation of Signaturizers, which are now stereochemically-aware, and provide an even more faithful description of complex small molecule bioactivity properties.

bioinformatics↗

bioRxiv2023

BQsupports: systematic assessment of the support and novelty of new biomedical associations

Living a Big Data era in Biomedicine, there is an unmet need to systematically assess experimental observations in the context of available information. This assessment would offer a means for an unbiased validation of the results and provide an initial estimate of the potential novelty of the findings. Here we present BQsupports, a web-based tool built upon the Bioteque biomedical descriptors that systematically analyzes and quantifies the current support to a given set of observations. The tool relies on over 1,000 distinct types of biomedical descriptors, covering over 11 different biological and chemical entities, including genes, cell lines, diseases and small molecules. By exploring hundreds of descriptors, BQsupports provide support scores for each observation across a wide variety of biomedical contexts. These scores are then aggregated to summarize the biomedical support of the assessed dataset as a whole. Finally, the BQsupports also suggests predictive features of the given dataset, which can be exploited in downstream machine learning applications. Availability and implementationThe web application and underlying data are available online (https://bqsuppports.irbbarcelona.org). Contactpatrick.aloy@irbbarcelona.org

bioinformatics↗

bioRxiv2022

Co-occurrence of essential gene dispensability and bypass suppressor mutations across species

Genes have been historically classified as either essential or non-essential based on their requirement for viability. However, some genes are essential in some genetic backgrounds but non-essential in others, thus challenging the binary classification of gene essentiality. Such dispensable essential genes represent a valuable model for understanding the incomplete penetrance of loss-of-function mutations that is often observed in natural populations. Here, we compiled data from multiple studies on essential gene dispensability in Saccharomyces cerevisiae to comprehensively characterize these genes. In analyses spanning different evolutionary time-scales, ranging from S. cerevisiae strains to human cell lines, dispensable essential genes exhibited distinct phylogenetic properties compared to other essential and non-essential genes. Integration of interactions with suppressor genes that can bypass the gene essentiality revealed the high functional modularity of the bypass suppression network. Furthermore, dispensable essential and bypass suppressor gene pairs reflected simultaneous changes in the mutational landscape of S. cerevisiae strains. Importantly, species in which dispensable essential genes were non-essential tended to carry bypass suppressor mutations in their genomes. Overall, our study offers a comprehensive view of dispensable essential genes and illustrates how their interactions with bypass suppressor genes reflect evolutionary outcomes.

genetics↗

bioRxiv2022

Global analysis of suppressor mutations that rescue human genetic defects

Genetic suppression occurs when the deleterious effects of a primary "query" mutation, such as a disease-causing mutation, are rescued by a suppressor mutation elsewhere in the genome. To capture existing knowledge on suppression relationships between human genes, we examined 2,400 published papers for potential interactions identified through either genetic modification of cultured human cells or through association studies in patients. The resulting network encompassed 476 unique suppression interactions that frequently linked genes that function in the same biological process. Suppressors were strongly enriched for genes with a role in stress response or signaling, suggesting that deleterious mutations can often be buffered by modulating signaling cascades or immune responses. Suppressor mutations tended to be deleterious when they occurred in absence of the query mutation, in apparent contrast with their protective role in the presence of the query. We formulated and quantified mechanisms of genetic suppression that could explain 71% of interactions and provided mechanistic insight into disease pathology. Finally, we used these observations to accurately predict suppressor genes in the human genome. The emerging frequency of suppression interactions and range of underlying mechanisms suggest that compensatory mutations may exist for the majority of genetic diseases.

genomics↗

bioRxiv2022

Characterization of p38a signaling networks in cancer cells using quantitative proteomics and phosphoproteomics

p38 (encoded by MAPK14) is a protein kinase that regulates cellular responses to almost all types of environmental and intracellular stresses. Upon activation, p38 phosphorylates many substrates both in the cytoplasm and nucleus, allowing this pathway to regulate a wide variety of cellular processes. While the role of p38 in the stress response has been widely investigated, its implication in cell homeostasis is less understood. To investigate the signaling networks regulated by p38 in normally proliferating cancer cells, we performed quantitative proteomic and phosphoproteomic analyses in breast cancer cells in which this pathway had been either genetically targeted or chemically inhibited. Our study identified with high confidence 35 proteins and 82 phosphoproteins (114 phosphosites) that are modulated by p38, and highlighted the implication of various protein kinases, including MK2 and mTOR, in the p38-regulated signaling networks. Moreover, functional analyses revealed an important contribution of p38 to the regulation of cell adhesion, DNA replication and RNA metabolism. Indeed, we provide experimental evidence supporting that p38 negatively regulates cell adhesion, and showed that this p38 function is likely mediated by the modulation of the adaptor protein ArgBP2. Collectively, our results illustrate the complexity of the p38 regulated signaling networks, provide valuable information on p38-dependent phosphorylation events in cancer cells, and document a mechanism by which p38 can regulate cell adhesion.

biochemistry↗

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

Binary Interactome Models of Inner- Versus Outer-Complexome Organization

Hundreds of different protein complexes that perform important functions across all cellular processes, collectively comprising the "complexome" of an organism, have been identified1. However, less is known about the fraction of the interactome that exists outside the complexome, in the "outer-complexome". To investigate features of "inner"- versus outer-complexome organisation in yeast, we generated a high-quality atlas of binary protein-protein interactions (PPIs), combining three previous maps2-4 and a new reference all-by-all binary interactome map. A greater proportion of interactions in our map are in the outer-complexome, in comparison to those found by affinity purification followed by mass spectrometry5-7 or in literature curated datasets8-11. In addition, recent advances in deep learning predictions of PPI structures12 mirror the existing experimentally resolved structures in being largely focused on the inner complexome and missing most interactions in the outer-complexome. Our new PPI network suggests that the outer-complexome contains considerably more PPIs than the inner-complexome, and integration with functional similarity networks13-15 reveals that interactions in the inner-complexome are highly detectable and correspond to pairs of proteins with high functional similarity, while proteins connected by more transient, harder-to-detect interactions in the outer-complexome, exhibit higher functional heterogeneity.

systems biology↗

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

Diverse natural variants suppress mutations in hundreds of essential genes

The consequence of a mutation can be influenced by the context in which it operates. For example, loss of gene function may be tolerated in one genetic background, but lead to lethality in another. The extent to which mutant phenotypes are malleable, the complexity of the architecture of modifiers, and the identities of causal genes and pathways remain largely unknown. Here, we measure the fitness effects of ~1,500 temperature sensitive alleles of yeast essential genes in the context of variation from ten different natural genetic backgrounds, and map the modifiers for 19 combinations. Altogether, fitness defects for 183 of the 530 tested genes (35%) could be suppressed by standing genetic variation in at least one wild strain. Suppression was generally driven by gain-of-function of a single, strong modifier gene. The validated causes included both variants in protein interaction partners or pathway members suppressing specific genes, as well as general modifiers altering the effect of many temperature sensitive alleles. The emerging frequency of suppression and range of possible suppression mechanisms suggest that a substantial fraction of monogenic diseases could be repressed by modulating other gene products.

genetics↗