Karim Beguir

A multispecies foundation model for genome function prediction, gene annotation and sequence design

NTv3 is a foundation model that predicts genome function, annotates genes, and designs DNA sequences across species, outperforming existing models in multiple tasks.

Annotating the genome at single-nucleotide resolution with DNA foundation models

The paper introduces SegmentNT, a DNA foundation model for genome annotation, achieving state-of-the-art performance in identifying 14 genic and regulatory elements across species.

De novo peptide sequencing with InstaNovo: Accurate, database-free peptide identification for large scale proteomics experiments

InstaNovo is a transformer model for de novo peptide sequencing that outperforms existing methods, enabling accurate identification of peptides and discovery of novel sequences and organisms.

The Nucleotide Transformer: Building and Evaluating Robust Foundation Models for Human Genomics

The paper introduces the Nucleotide Transformer, a foundation model for genomics, showing it can predict molecular phenotypes from DNA sequences with transfer learning across tasks and organisms.

A Foundational Large Language Model for Edible Plant Genomes

This paper introduces a large language model for plant genomics that provides state-of-the-art predictions for regulatory annotations and variant prioritization, along with a benchmark dataset for deep learning methods.

ChatNT: A Multimodal Conversational Agent for DNA, RNA and Protein Tasks

ChatNT is a conversational agent that understands biological sequences and achieves state-of-the-art results on genomics tasks while being accessible to non-coders.

LightMHC: A Light Model for pMHC Structure Prediction with Graph Neural Networks

LightMHC predicts full-atom pMHC structures from sequences using a lightweight model with attention mechanisms, achieving comparable performance to AlphaFold 2 and ESMFold with faster speed.

Early Computational Detection of Potential High Risk SARS-CoV-2 Variants

A COVID-19 Early Warning System combines structural modelling with machine learning to detect and monitor high risk SARS-CoV-2 variants, identifying all 16 WHO designated variants on average more than one and a half months in advance by selecting on average less than 0.3% of the weekly novel variants.

Peptide-MHC Structure Prediction With Mixed Residue and Atom Graph Neural Network

The paper introduces a graph neural network for peptide-MHC structure prediction, achieving accuracy comparable to AlphaFold 2 with significantly fewer parameters.

So ManyFolds, So Little Time: Efficient Protein Structure Prediction With pLMs and MSAs

The paper presents a streamlined AlphaFold variant using pLMs and MSAs, showing that combining or using either modality achieves comparable performance to models trained exclusively on one.