Píetro Lió

Interpreting Graph Neural Networks for Antibacterial Drug Repurposing against Acinetobacter baumannii

This paper uses a graph neural network to predict antibacterial activity against Acinetobacter baumannii and identifies piperazine as a key contributor to predicted activity, leading to the testing of CE3F4.

LLMsFold: Integrating Large Language Models and Biophysical Simulations for De Novo Drug Design

LLMsFold designs and validates new small molecules targeting pathogenic proteins using LLMs and biophysical simulations, showing strong in silico binding and drug-like profiles.

Five hundred million years of methylation: tracing the mutational origins of vertebrate genome composition

The paper analyzes genome composition across vertebrates and invertebrates to show that CpG depletion defines a compositional axis, with AG/CT as its strongest marker, not the direct mutational product TpG/CpA.

FibrilNet maps conserved and tissue-specific molecular environments across systemic amyloidoses

The paper introduces FibrilNet, a network framework that identifies conserved and tissue-specific molecular environments across systemic amyloidoses by integrating proteomic data with semantic and interaction information.

De novo design of RNA pseudoknots with deep learning

This paper demonstrates that deep learning can accurately design RNA pseudoknots, achieving results comparable to human experts in a competition.

Deep learning enables cross-species annotation and attribution of ageing states in haematopoietic stem and immune cells

The paper develops a cross-species deep learning framework to annotate and attribute ageing states in haematopoietic stem and immune cells, achieving high AUROC scores in human data.

The illusion of interpretability in biologically informed neural networks

This paper shows that biologically informed neural networks (BINNs) cannot reliably recover internal biological structures from outputs alone due to nonidentifiability.

Topography Aware Optimal Transport for Alignment of Spatial Omics Data

TOAST improves spatial omics data alignment by integrating spatial constraints into optimal transport, enhancing accuracy and biological interpretability.

Multiplexed, scalable analog recording of gene regulation dynamics over weeks using intracellular protein tapes

This paper introduces CytoTape, a tool for continuous, multiplexed recording of gene regulation dynamics over weeks at single-cell resolution, and demonstrates its use in measuring transcription factor activities.

Generative inverse design of RNA structure and function with gRNAde

This paper introduces gRNAde, an RNA language model for inverse design of RNA structure and function, achieving success rates comparable to human experts in designing pseudoknotted RNAs.

Nanoplasmonic SERS-enabled structural identification of an uncharacterised indole metabolite directly in E. coli cultures

This study identifies a novel indole metabolite, I*, produced by TnaA in E. coli using nanoplasmonic SERS, without isolation or purification.

Uncertainty-Aware Tau Detection in Progressive Supranuclear Palsy Using Object Detection Models

This paper uses deep learning object detection models to automate tau lesion identification in PSP brain slides, incorporating uncertainty quantification to improve reliability and clinical relevance.

Integration of Bioinformatics and Machine Learning to Characterize Fusobacterium nucleatum's Pathogenicity

This study identifies a candidate pathogenicity island in Fusobacterium nucleatum using bioinformatics and machine learning, linking iron acquisition to cancer promotion.

Structure-based Generation of a Secondary Nucleation Inhibitor in alpha-Synuclein Aggregation Using a Conditional Diffusion Model

The paper uses a conditional diffusion model to design a small molecule inhibitor of secondary nucleation in alpha-synuclein aggregation, showing it effectively inhibits fibril proliferation.

Accelerating Biomolecular Modeling with AtomWorks andRF3

AtomWorks, a modular atom-level data framework, trains structure prediction and design models; the resulting RF3 improves chirality handling and performs between AF3 and Boltz.

GhostBuster: A Deep-Learning-based, Literature-Unbiased Gene Prioritization Tool for Gene Annotation Prediction

GhostBuster predicts gene functions and annotations while minimizing literature bias, outperforming biased datasets in identifying novel gene annotations.

Proteomizer: Leveraging the Transcriptome-Proteome Mismatch to Infer Novel Gene Regulatory Relations

Proteomizer infers proteomic profiles from transcriptomic and miRNomic data, improving differential expression analysis accuracy but with limited generalizability.

UnifiedGreatMod: A New Holistic Modeling Paradigm for Studying Biological Systems on a Complete and Harmonious Scale

The paper introduces a new modeling paradigm, UnifiedGreatMod, that integrates fine-grained and coarse-grained biological information to study system behavior, demonstrating its application in simulating E. coli metabolism and analyzing cell responses to infection.

Genome-wide DNA methylation study reveals specific signatures in the affected arterial tissue of giant cell arteritis patients

This study identifies DNA methylation signatures in GCA-affected arterial tissue, revealing disrupted inflammatory pathways and T cell exhaustion markers.

Learning and actioning general principles of cancer cell drug sensitivity

The paper develops interpretable models to predict cancer drug sensitivity using transcriptomics data and enhances predictions with large language models, validated on clinical samples.

Interaction between gene expression and morphokinetic parameters in undisturbed human embryo culture

This study compares gene expression in human embryos cultured in time-lapse systems versus conventional incubators and identifies molecular mechanisms linking morphokinetic parameters to embryo development.

MISATO - Machine learning dataset of protein-ligand complexes for structure-based drug discovery

The paper introduces MISATO, a dataset combining quantum mechanics properties and molecular dynamics simulations of protein-ligand complexes to support structure-based drug discovery.

Microdroplet screening rapidly profiles a biocatalyst to enable its AI-assisted engineering

The paper combines microfluidic screening with machine learning to rapidly engineer biocatalysts, achieving 23-fold improvement in catalytic efficiency through sequence-function mapping.

gRNAde: Geometric Deep Learning for 3D RNA inverse design

gRNAde designs RNA sequences based on 3D backbone structures using geometric deep learning, achieving higher native sequence recovery than Rosetta on benchmark tasks.

Hypergraph factorisation for multi-tissue gene expression imputation

HYFA improves multi-tissue gene expression imputation by leveraging hypergraphs and transfer learning, outperforming existing methods and enabling cell-type signature resolution from bulk data.

Spatio-relational inductive biases in spatial cell-type deconvolution

This paper introduces GNN-C2L, a spatial cell-type deconvolution method that improves performance by incorporating spatial inductive biases, outperforming spatial-agnostic variants on simulated datasets.

ElegansNet: a brief scientific report and initial experiments.

ElegansNet uses the C. elegans connectome to design deep learning systems that outperform random and artificial networks in image classification and digit reconstruction tasks.

A Deep Graph Neural Network Architecture for Modelling Spatio-temporal Dynamics in resting-state functional MRI Data

This paper introduces a deep graph neural network combining GNNs and TCNs to model spatio-temporal dynamics in rs-fMRI data, showing explainable neurobiological insights.

Decoding Surface Fingerprints for Protein-Ligand Interactions

The paper introduces a computational framework for fragment-based screening and binding affinity prediction using protein surface descriptors to identify ligands for challenging protein targets.

CellVGAE: An unsupervised scRNA-seq analysis workflow with graph attention networks

CellVGAE uses graph attention networks for scRNA-seq analysis, improving dimensionality reduction and clustering across multiple datasets.