Biology subjects

De Baets, B.

Publications and source records attributed to De Baets, B..

5 recordsLinked to original sources

bioRxiv2025

PhaLP 2.0: extending the community-oriented phage lysin database with a SUBLYME pipeline for metagenomic discovery

As biology becomes increasingly data-driven, so too does the field of phage lysins, enzymes that degrade bacterial cell walls and hold promise as alternatives to traditional antibiotics. Five years ago, we introduced PhaLP, a centralized resource for Phage Lytic Protein sequences and associated metadata to support global research efforts. Here, we present PhaLP 2.0, a significantly enhanced database designed to overcome key challenges in the computational study of lysins by integrating newly identified lysins obtained from thousands of metagenomes. To expand the known diversity of lysins beyond those from cultured phages, we developed SUBLYME, a protein embedding-based machine learning Software designed to Uncover and classify Bacteriophage Lysins in Metagenomic datasets. Using embeddings derived from the prior well-curated protein sequences of the original PhaLP database, we trained support vector machines to distinguish lysins from non-lysins in viromes and classify them as either endolysins or virion-associated lysins. The models achieved an average F1-score of 98% on held-out lysin clusters. SUBLYME enabled the discovery of 743,000 new lysin sequences from EnVhogDB, a virome-derived protein database, increasing the number of known lysin clusters by a factor of 40, from 1,000 to 40,000. PhaLP 2.0 entries were annotated by integrating Pfam functional predictions to the refined delineations obtained with SPAED, an algorithm that leverages the predicted aligned error matrix from AlphaFold predictions to identify domain boundaries. Both SUBLYME and the PhaLP 2.0 database are accessible online at https://github.com/Rousseau-Team/sublyme and http://phalp.ugent.be, respectively. Together, these advances establish PhaLP 2.0 as a comprehensive and scalable portal for the discovery, classification, and sequence analysis of phage lysins, paving the way for future antibacterial applications and evolutionary insights. O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=59 SRC="FIGDIR/small/692814v1_ufig1.gif" ALT="Figure 1"> View larger version (21K): org.highwire.dtl.DTLVardef@181e26borg.highwire.dtl.DTLVardef@37e099org.highwire.dtl.DTLVardef@7a7fccorg.highwire.dtl.DTLVardef@5c42f1_HPS_FORMAT_FIGEXP M_FIG C_FIG

bioinformatics↗

bioRxiv2024

DepoScope: accurate phage depolymerase annotation and domain delineation using large language models

Bacteriophages (phages) are viruses that infect bacteria. Many of them produce specific enzymes called depolymerases to break down external polysaccharide structures. Accurate annotation and domain identification of these depolymerases are challenging due to their inherent sequence diversity. Hence, we present DepoScope, a machine learning tool that combines a fine-tuned ESM-2 model with a convolutional neural network to precisely identify depolymerase sequences and their enzymatic domains. To accomplish this, we curated a dataset from the INPHARED phage genome database, created a polysaccharide-degrading domain database, and applied sequential filters to construct a high-quality dataset, which are subsequently used to train DepoScope. Our work is the first approach that combines sequence-level predictions with amino-acid-level predictions for an accurate depolymerase detection and functional domain identification. In that way, we believe that DepoScope can enhance our understanding of phage-host interactions at the level of depolymerases. Summary with Key MessagesO_LIPhage depolymerases are proteins that play a crucial role in the first step of a phage replication cycle. As a result, they are both important from a biological perspective and a therapeutical perspective. C_LIO_LICurrent methods to accurately annotate phage depolymerases and their associated enzymatic domains remains challenging due to their inherent high sequence diversity. C_LIO_LIWe have developed DepoScope, a language-based artificial intelligence model that can accurately identify phage depolymerases and their specific enzymatic domains. C_LIO_LIWe provide full public access to the DepoScope code and database to give broad access to the research community and promote further research. C_LI

bioinformatics↗

bioRxiv2023

A Comparison of Embedding Aggregation Strategies in Drug-Target Interaction Prediction

The prediction of interactions between novel drugs and biological targets is a vital step in the early stage of the drug discovery pipeline. Many deep learning approaches have been proposed over the last decade, with a substantial fraction of them sharing the same underlying two-branch architecture. Their distinction is limited to the use of different types of feature representations and branches (multi-layer perceptrons, convolutional neural networks, graph neural networks and transformers). In contrast, the strategy used to combine the outputs (embeddings) of the branches has remained mostly the same. The same general architecture has also been used extensively in the area of recommender systems, where the choice of an aggregation strategy is still an open question. In this work, we investigate the effectiveness of three different embedding aggregation strategies in the area of drug-target interaction (DTI) prediction. We formally define these strategies and prove their universal approximator capabilities. We then present experiments that compare the different strategies on benchmark datasets from the area of DTI prediction, showcasing conditions under which specific strategies could be the obvious choice.

bioinformatics↗

bioRxiv2021

BioCCP.jl: Collecting Coupons in combinatorial biotechnology

In combinatorial biotechnology, it is crucial for screening experiments to sufficiently cover the design space. In the BioCCP.jl package, we provide functions for minimum sample size determination based on the mathematical framework coined the Coupon Collector Problem. BioCCP.jl, including source code, documentation and a Pluto notebook is available at https://github.com/kirstvh/BioCCP.

bioengineering↗

bioRxiv2018

Algebraic Shortcuts for Leave-One-Out Cross-Validation in Supervised Network Inference

MotivationSupervised machine learning techniques have traditionally been very successful at reconstructing biological networks, such as protein-ligand interaction, protein-protein interaction and gene regulatory networks. Recently, much emphasis has been placed on the correct evaluation of such supervised models. It is vital to distinguish between using the model to either predict new interactions in a given network or to predict interactions for a new vertex not present in the original network. Specific cross-validation schemes need to be used to assess the performance in such different prediction settings.\n\nResultsWe present a series of leave-one-out cross-validation shortcuts to rapidly estimate the performance of state-of-the-art kernel-based network inference techniques.\n\nAvailabilityThe machine learning techniques with the algebraic shortcuts are implemented in the RLScore software package.

systems biology↗