Élodie Laine

The paper evaluates how different constructed negative examples affect sequence-based protein-protein interaction models, showing significant variability in performance metrics based on negative construction methods.

phyloDS identifies differentially used orthologous splice junctions across species by improving orthology definition, statistical testing, and sequence-level validation.

ProteoCast predicts the functional impact of missense variants using evolutionary information, achieving high accuracy across the Drosophila proteome.

ProteoCast predicts and interprets missense variant effects using evolutionary and structural analysis, achieving high accuracy in pathogenicity prediction and identifying functional motifs in disordered regions.

SeaMoon predicts continuous protein motions from sequences using pLM embeddings and a lightweight CNN, achieving reasonable accuracy for 40% of test proteins.

VespaG is a fast missense variant effect predictor using protein Language Models, achieving high accuracy on the ProteinGym benchmark.

This paper introduces DANCE, a method to analyze conformational variability in protein families, and evaluates dimensionality reduction techniques for sampling unseen protein states.

The paper predicts mutational outcomes for 1.5M missense variants across 72 protein families using alignment-based methods, showing high-quality and compute-efficient results for entire proteomes.

The paper introduces ASPRING, a method to detect alternatively spliced protein repeats across species, identifies over 5,000 conserved repeats, and maps them to 3D structures to study their role in protein interactions.

The paper identifies N-terminal signatures in yeast proteins that influence mitochondrial import, showing hydrophobic residues at position 2 are critical and conserved across Saccharomycotina yeasts.

DLA estimates binding affinity changes upon mutation by analyzing protein interfaces with 3D convolutions and achieves a Pearson correlation of 0.81 on 2000 mutations.

DLA-mutation predicts mutation-induced changes in protein-protein binding affinity using a deep learning framework trained on protein complex structures.

DLA-Ranker evaluates protein docking conformations using locally oriented cubes to distinguish near-native from incorrect ones, outperforming other deep learning-based scoring functions.

This paper introduces a molecular cross-docking approach to identify protein partners and assesses its performance against sequence-based deep learning methods.

The paper introduces a deep learning method to rank protein docking conformations by analyzing local interface geometry and combines it with physics-based and statistical scoring functions.

The paper develops evolutionary splicing graphs to assess alternative splicing conservation across species and identifies genes where alternative splicing affects pseudo-repeats.

LEVELNET is a tool for visualizing, exploring, and comparing protein-protein interaction networks using multilayered graphs and structural data from the Protein Data Bank.

ProfileView uses probabilistic models to classify protein sequences and resolve functional diversity across different families, outperforming existing methods.

HOPMA predicts protein functional states and transitions using a modified elastic network model and contact maps, improving conformational space exploration.

This study develops a computational framework to quantify disorder in coiled-coils and design substitutions to modulate their stability and flexibility.

The paper introduces a method to predict protein conformational transitions using nonlinear twists from normal mode analysis, enabling access to previously inaccessible motions while preserving structure.

The paper reconstructs the evolutionary history of JNK family transcripts and models their protein structures to understand how alternative splicing and transcription sites contribute to functional diversity.

The paper introduces a method to predict protein-DNA interfaces using evolutionary, physico-chemical, and geometric properties, achieving higher sensitivity than existing methods.

GEMME predicts mutational effects by modeling evolutionary history, outperforming existing methods in speed and accuracy for viral and conserved sequences.

The paper introduces methods to predict protein interaction regions by integrating evolutionary, physico-chemical, and geometrical data with docking simulations.

This paper introduces GEMME, a method for predicting mutational effects using sequence data and enhances it with structural and allele frequency data to improve biological insights.