Patrick Aloy

The paper adapts a pre-trained biological foundation model to learn perturbation-centric representations by fine-tuning it on perturbation profiles, improving classification and uncovering new biological relationships.

This paper designs cell-specific small molecule cytotoxics using AI, combining screening, machine learning, and reinforcement learning, resulting in validated compounds with improved hit discovery rates.

The paper identifies suppressor variants that alter gene essentiality across yeast strains, revealing context-dependent genetic mechanisms.

The paper identifies inhibitor-induced kinase destabilization as a common phenomenon and reveals that it occurs through endogenous proteolytic mechanisms rather than classical ubiquitination.

This study assesses how spatial microenvironments influence gene expression and cell-cell communication using scRNA-seq and spatial transcriptomics, finding limited transcriptional changes and weak correlation between communication predictions and spatial colocalization.

The paper introduces a protocol to integrate diverse bioactivity data into the Chemical Checker compound universe, enabling the generation of novel bioactivity signatures for small molecules.

The paper systematically studies hybrid triplex topology and stability, revealing a regulatory mechanism in the human genome and transcriptome through triplex formation.

The paper introduces PocketVec, a method to generate binding site descriptors for proteins, identifies over 32,000 druggable pockets in the human proteome, and shows its utility in discovering similar pockets not detected by traditional methods.

The paper trains stereochemically-aware models to better capture bioactivity differences between stereoisomers of small molecules.

The paper identifies suppressor mutations that rescue human genetic defects by analyzing 2,400 papers and predicting suppressor genes in the human genome.

BQsupports assesses the support and novelty of biomedical associations by analyzing descriptors across multiple contexts and providing scores for downstream applications.

The paper maps binary SARS-CoV-2 protein interactions with host proteins, identifying key processes like immune signaling and ubiquitination, and suggests therapeutic targets.

The paper introduces Bioteque, a large-scale biomedical knowledge graph resource that provides pre-calculated descriptors for downstream machine learning tasks.

The paper identifies molecular signatures of Alzheimer's disease in AD mouse models and finds drugs that reverse these signatures and cognitive impairment.

The paper measures how natural genetic variation can suppress mutations in essential yeast genes, finding that 35% of tested genes can be suppressed by modifier genes in different backgrounds.

The paper constructs a high-quality yeast binary interactome map and compares inner- and outer-complexome organization, finding the outer-complexome contains more interactions and exhibits higher functional heterogeneity.

The paper curates and expands the yeast complexome, showing increased genetic interactions and functional links among complex members.

Proximity-dependent biotinylation (BioID) of 28 SARS-CoV-2 proteins in human cells maps 5,415 proximal interactions with 2,128 host targets across basal and poly(I:C) conditions.

The paper introduces deep neural networks to infer bioactivity signatures for uncharacterized compounds, improving activity prediction across various biological endpoints.

The paper investigates the role of inherited and de novo truncating mutations in high functioning autism and identifies pleiotropic effects of LRP1 variants across psychiatric phenotypes.

The paper introduces Chemical Checker, a resource that integrates bioactivity data on 800,000 small molecules and demonstrates its use in drug discovery and disease modeling.

The paper presents HuRI, a comprehensive human protein-protein interaction map, and demonstrates its utility in understanding cellular function and disease mechanisms.

The paper develops a network-based method to identify gene modules associated with drug response in cancer cell lines, improving interpretability and reducing noise in transcriptomic signatures.

This paper develops a deep mutational scanning framework to map functional missense variants in human genes and applies it to four proteins, identifying pathogenic variants and showing scalability to disease genes.