Biology subjects

Mallet, V.

Publications and source records attributed to Mallet, V..

7 recordsLinked to original sources

bioRxiv2026

Learning Dynamic Protein Representations at Scale with Distograms

Protein function and other biological properties often depend on structural dynamics, yet most machine-learning predictors rely on static representations. Physics-based molecular simulations can describe conformational variability but remain computationally prohibitive at scale. Generative models provide a more efficient alternative, though their ability to produce accurate conformational ensembles is still limited. In this work, we bypass expensive simulations by leveraging residue-residue distance probability distributions (distograms) from structure predictors such as AlphaFold2. Our approach provides a scalable way to encode dynamic information into protein representations, aiming to improve function prediction without explicit conformational sampling.

bioinformatics↗

bioRxiv2023

Semi-supervised learning and large-scale docking data accelerate RNA virtual screening

RNAs constitute a vast reservoir of mostly untapped drug targets. Structure-based virtual screening (VS) methods screen large compound libraries for identifying promising candidate molecules by conditioning on binding site information. The classical approach relies on molecular docking simulations. However, this strategy does not scale well with the size of the small molecule databases and the number of potential RNA targets. Machine learning emerged as a promising technology to resolve this bottleneck. Efficient data-driven VS methods have already been introduced for proteins, but these techniques have not yet been developed for RNAs due to limited dataset sizes and lack of practical use-case evaluation. We propose a data-driven VS pipeline that deals with the unique challenges of RNA molecules through coarse grained modeling of 3D structures and heterogeneous training regimes using synthetic data augmentation and RNA-centric self supervision. We report strong prediction and generalizability of our framework, ranking active compounds among inactives in the top 2.8% on average on a structurally distinct drug-like test set. Those predictions are sensitive, but robust to pockets alterations, opening the door to its use on binding site detection methods outputs. Our model results in a ten thousand-times speedup over docking techniques while obtaining higher performance. Finally, we deploy our model on a recently published in-vitro small molecule microarray experiment with 20,000 compounds and report a mean enrichment factor at 1% of 2.93 on four unseen RNA riboswitches. To our knowledge, this is the first experimental evidence of success for structure-based deep learning methods in RNA virtual screening. Our source code and data, as well as a Google Colab notebook for inference, are available on GitHub.1

bioinformatics↗

bioRxiv2023

Finding Antibodies in Cryo-EM densities with CrAI

Therapeutic antibodies have emerged as a prominent class of new drugs due to their high specificity and their ability to bind to several protein targets. Once an initial antibody has been identified, an optimization of this hit compound follows based on the 3D structure, when available. Cryo-EM is currently the most efficient method to obtain such structures, supported by well-established methods that can transform raw data into a potentially noisy 3D map. These maps need to be further interpreted by inferring the number, position and structure of antibodies and other proteins that might be present. Unfortunately, existing automated methods addressing this last step have a limited accuracy and usually require additional inputs, high resolution maps, and exhibit long running times. We propose the first fully automatic and efficient method dedicated to finding antibodies in cryo-EM densities: CrAI. This machine learning approach leverages the conserved structure of antibodies and exploits a dedicated novel database that we built to solve this problem. Running a prediction takes only a few seconds, instead of hours, and requires nothing but the cryo-EM density, seamlessly integrating in automated analysis pipelines. Our method is able to find the location of both Fabs and VHHs, at resolutions up to 10[A] and is significantly more reliable than existing methods. It also provides an accurate estimation of the antibodies pose, even in challenging examples such as Fab binding to VHHs and vice-versa. We make our method available as a ChimeraX[44] bundle. 1

bioinformatics↗

bioRxiv2021

InDeep : 3D fully convolutional neural networks to assist in silico drug design on protein-protein interactions

AO_SCPLOWBSTRACTC_SCPLOWO_ST_ABSMotivationC_ST_ABSProtein-protein interactions (PPIs) are key elements in numerous biological pathways and the subject of a growing number of drug discovery projects including against infectious diseases. Designing drugs on PPI targets remains a difficult task and requires extensive efforts to qualify a given interaction as an eligible target. To this end, besides the evident need to determine the role of PPIs in disease-associated pathways and their experimental characterization as therapeutics targets, prediction of their capacity to be bound by other protein partners or modulated by future drugs is of primary importance. ResultsWe present InDeep, a tool for predicting functional binding sites within proteins that could either host protein epitopes or future drugs. Leveraging deep learning on a curated data set of PPIs, this tool can proceed to enhanced functional binding site predictions either on experimental structures or along molecular dynamics trajectories. The benchmark of InDeep demonstrates that our tool outperforms state of the art ligandable binding sites predictors when assessing PPI targets but also conventional targets. This offers new opportunities to assist drug design projects on PPIs by identifying pertinent binding pockets at or in the vicinity of PPI interfaces. AvailabilityThe tool is available on GitHub3 along with a PyMol plugin for visualization. Predictions of InDeep can be consulted at iPPI-DB4

bioinformatics↗

bioRxiv2021

Reverse-Complement Equivariant Networks for DNA Sequences

As DNA sequencing technologies keep improving in scale and cost, there is a growing need to develop machine learning models to analyze DNA sequences, e.g., to decipher regulatory signals from DNA fragments bound by a particular protein of interest. As a double helix made of two complementary strands, a DNA fragment can be sequenced as two equivalent, so-called Reverse Complement (RC) sequences of nucleotides. To take into account this inherent symmetry of the data in machine learning models can facilitate learning. In this sense, several authors have recently proposed particular RC-equivariant convolutional neural networks (CNNs). However, it remains unknown whether other RC-equivariant architectures exist, which could potentially increase the set of basic models adapted to DNA sequences for practitioners. Here, we close this gap by characterizing the set of all linear RC-equivariant layers, and show in particular that new architectures exist beyond the ones already explored. We further discuss RC-equivariant pointwise nonlinearities adapted to different architectures, as well as RC-equivariant embeddings of k-mers as an alternative to one-hot encoding of nucleotides. We show experimentally that the new architectures can outperform existing ones.

bioinformatics↗

bioRxiv2020

OptiMol: Optimization of binding affinities in chemical space for drug discovery

Ligand-based drug design has recently benefited from the boost of deep generative models. These models enable extensive explorations of the chemical space, and provide a platform for molecular optimization. However, current state of the art methods do not leverage the structure of the target, which is known to play a key role in the interaction. We propose an optimization pipeline that leverages complementary structure-based and ligand-based methods. Instead of performing docking on a fixed drug bank, we iteratively select promising compounds in the whole chemical space using a ligand-centered generative model. Molecular docking is then used as an oracle to guide compound optimization. This allows to iteratively generate leads that better fit the target structure, in a closed optimization loop, without prior knowledge about bio-actives. For this purpose, we introduce a new graph to selfies VAE which benefits from a seventeen times faster decoding than graph to graph methods while being competitive with the state of the art. We then successfully optimize the generation of molecules towards high docking scores, enabling a ten-fold augmentation of high-scoring compounds found with a fixed computational budget. AvailabilityCode is available on GitHub Contactjeromew@mcgill.ca

bioinformatics↗

bioRxiv2019

Extended RNA base pairing networks imprint small molecule binding preferences

AO_SCPLOWBSTRACTC_SCPLOWO_ST_ABSMotivationC_ST_ABSThe binding of small molecules to RNAs is an important mechanism which can stabilize 3D structures or activate key molecular functions. To date, computational and experimental efforts toward small molecule binding prediction have primarily focused on protein targets. Considering that a very large portion of the genome is transcribed into non-coding RNAs but only few regions are translated into proteins, successful annotations of RNA elements targeted by small-molecule would likely uncover a vast repertoire of biological pathways and possibly lead to new therapeutic avenues. ResultsOur work is a first attempt at bringing machine learning approaches to the problem of RNA drug discovery. RNAmigos takes advantage of the unique structural properties of RNA to predict small molecule ligands for unseen binding sites. A key feature of our model is an efficient representation of binding sites as augmented base pairing networks (ABPNs) aimed at encoding important structural patterns. We subject our ligand predictions to two virtual screen settings and show that we are able to rank the known ligand on average in the 73rd percentile, showing a significant improvement over several baselines. Furthermore, we observe that graphs which are augmented with non-Watson Crick (a.k.a non-canonical) base pairs are the only representation which is able to retrieve a significant signal, suggesting that non-canonical interactions are an necessary source of binding specificity in RNAs. We also find that an auxiliary graph representation task significantly boosts performance by providing efficient structural embeddings to the low data setting of ligand prediction. RNAmigos shows that RNA binding data contains structural patterns with potential for drug discovery, and provides methodological insights which can be applied to other structure-function learning tasks. AvailabilityCode and data is freely available at http://csb.cs.mcgill.ca/RNAmigos. Contactjerome@cs.mcgill.ca

bioinformatics↗