Martin F. Larralde

The paper develops a geometric matching method to detect catalytic features in protein structures, showing it is more sensitive than sequence-based approaches for identifying homology between distantly related enzymes.

This study identifies two novel biosynthetic intermediates and characterizes their enzymes in the unique glycan modification pathway of C. difficile flagellin.

The paper analyzes the biosynthetic potential of the Bacillus cereus group using genomic data, identifying many novel biosynthetic gene clusters.

The paper introduces ontology pre-training to improve machine learning predictions for metabolites by integrating bio-ontology knowledge, showing better performance and faster training.

Sismis is a machine learning tool that accurately detects and classifies secretion systems in bacterial (meta)genomes, identifying many previously unknown systems.

The paper introduces Python libraries for computing average nucleotide identity, matching existing tools' results and integrating with BioPython for workflow use.

IGUA is a fast and scalable method for delineating gene cluster families in prokaryotic genomes, outperforming existing tools in speed without sacrificing accuracy.

CHAMOIS predicts metabolite properties from biosynthetic gene clusters using machine learning, achieving AUPRC > 0.5 for 120 chemical properties.

The paper surveys Ruminococcus gnavus distribution across host lifestyles and health, revealing higher prevalence in Westernized populations and genomic differences between healthy and Crohn's disease isolates.

This study shows that long-term fasting changes gut microbial composition and function, as well as the serum metabolome, with significant shifts in microbial species and metabolite levels.

The paper predicts stop codon reassignment in bacteriophage genomes to improve functional annotation, showing significant increases in gene length and coding density.

The paper introduces the ISA API, a Python library for handling ISA-Tab and ISA-JSON metadata formats, enabling programmatic manipulation and interoperability in life science experiments.

A linear conditional random field over ordered protein-domain annotations finds biosynthetic gene clusters de novo more precisely than DeepBGC and runs over three times faster on one CPU.