PhageLysClass: Evidence-aware classification of phage lytic proteins using protein language models

Phage lytic proteins are promising antibacterial and biotechnological candidates, yet uncertain annotations, modular architectures, and class imbalance complicate sequence-based classification. We developed PhageLysClass, an evidence-aware framework for recognizing lytic proteins and distinguishing Endolysins, Depolymerases, and virion-associated lytic enzymes. From 442,329 unique phage sequences, including 11,867 evidence-supported positives, we defined distinct candidate non-target, unlabeled, and challenge populations. Seven supervised and two positive-unlabeled tasks were evaluated across factorial redundancy and train-test homology conditions, yielding 2,240 and 768 experiments, respectively. Redundancy reduction influenced generalization more strongly than homology separation, with non-additive interactions between them. ESM2-T30 consistently outperformed alternative representations and supported a hierarchical classifier achieving approximately 0.89 MCC under strict R30/S30 conditions in the selected, pre-refitting benchmark. Applied without retraining to 77,897 sequences across three collections, PhageLysClass combined predictions with physicochemical and geometric contextualization, highlighting Depolymerase-like candidates among structurally ambiguous proteins while distinguishing reference-supported assignments from extrapolation.

Authors

Institutions

Publication Details

Journal
bioRxiv (Cold Spring Harbor Laboratory)
Published
2026-11-20
DOI
https://doi.org/10.64898/2026.11.19.767946
Primary Topic
Bacteriophages and microbial interactions
Type
preprint

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

PhageLysClass: Evidence-aware classification of phage lytic proteins using protein language models

Alejandro Villasante, María Elena Lienqueo, David Medina-Ortiz, Álvaro Olivera‐Nappa et al.
4 links
bioRxiv (Cold Spring Harbor Laboratory)
Bacteriophages and microbial interactions
preprint

PhageLysClass: Evidence-aware classification of phage lytic proteins using protein language models

Alejandro Villasante, María Elena Lienqueo, David Medina-Ortiz, Álvaro Olivera‐Nappa, Jaime Romero
preprint en

Abstract

Phage lytic proteins are promising antibacterial and biotechnological candidates, yet uncertain annotations, modular architectures, and class imbalance complicate sequence-based classification. We developed PhageLysClass, an evidence-aware framework for recognizing lytic proteins and distinguishing Endolysins, Depolymerases, and virion-associated lytic enzymes. From 442,329 unique phage sequences, including 11,867 evidence-supported positives, we defined distinct candidate non-target, unlabeled, and challenge populations. Seven supervised and two positive-unlabeled tasks were evaluated across factorial redundancy and train-test homology conditions, yielding 2,240 and 768 experiments, respectively. Redundancy reduction influenced generalization more strongly than homology separation, with non-additive interactions between them. ESM2-T30 consistently outperformed alternative representations and supported a hierarchical classifier achieving approximately 0.89 MCC under strict R30/S30 conditions in the selected, pre-refitting benchmark. Applied without retraining to 77,897 sequences across three collections, PhageLysClass combined predictions with physicochemical and geometric contextualization, highlighting Depolymerase-like candidates among structurally ambiguous proteins while distinguishing reference-supported assignments from extrapolation.

bioRxiv (Cold Spring Harbor Laboratory)
Centre for Biotechnology and Bioengineering (CL), Universidad de Magallanes (CL), Universidad de Chile (CL)
Agencia Nacional de Investigación y Desarrollo
Bacteriophages and microbial interactions
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon
PhageLysClass: Evidence-aware classification of phage lytic proteins using protein language models — Alejandro Villasante, María Elena Lienqueo, et al. · bioRxiv (Cold Spring Harbor Laboratory) (2026) | TGRS Research Map | TGRS