SelCysML: a hybrid ML-BLAST pipeline for the automated identification and functional annotation of bacterial selenoproteins
- 1Applied Mathematics Department, NEIKER – Basque Institute for Agricultural Research and Development, Basque Research and Technology Alliance (BRTA), Derio, Bizkaia, Spain
- 2Genetics, Genomics and Microbiology Research Group, Institute for Multidisciplinary Research in Applied Biology (IMAB), Universidad Pública de Navarra (UPNA), Pamplona, Spain
Abstract
Selenoproteins are characterized by the incorporation of selenocysteine (Sec) via UGA codon recoding, play vital roles in redox homeostasis and enzymatic catalysis. Despite this biological significance, their computational identification remains challenging, and recent assessments confirm that public databases suffer from widespread annotation errors, often misclassifying these proteins. Current predictive tools frequently operate as “black boxes” with limited interpretability or lack integrated functional validation layers, which contributes to high false-positive rates and persistent ambiguities in the annotation of the selenoproteome. We present SelCysML, an accessible and highly accurate computational pipeline that integrates an ensemble of machine learning (ML) models with a parallel functional validation module using BLAST. By utilizing curated physicochemical descriptors—Amino Acid Composition (AAC) and Composition, Transition, Distribution (CTDC)—the model achieves a robust performance, with an AUC-ROC of 0.968 and a Matthews Correlation Coefficient (MCC) of 0.787. SelCysML not only classifies protein sequences but also provides functional evidence through sequence similarity, effectively bridging the gap between automated prediction and experimental biological insight. SelCysML provides a user-friendly, interpretable, and precise solution for the identification of bacterial selenoproteins in proteomic datasets, serving as a reliable tool for both bioinformaticians and experimental biologists.
Related
- Identification and bioinformatic functional annotation of circRNAs mediating phenotypic differences in hair follicle cycling of Jinlan cashmere goats
- NMRSCOUT: Automated Annotation of Complex Mixtures from Multidimensional NMR Spectra Using Predictive Modelling and Statistical Filtering
- From functional annotation to functional meaning in microbiome research
- EchoGO: annotation-context-aware functional interpretation of transcriptomic data.
- Functional annotation of insulin-responsive genes in human adipocytes reveals PLCXD1 as a lipid storage regulator
BibTeX
@article{lavin2026selcysml,
title = {{SelCysML: a hybrid ML-BLAST pipeline for the automated identification and functional annotation of bacterial selenoproteins}},
author = {José L. Lavín and José A. Oguiza},
journal = {Frontiers in Bioinformatics},
year = {2026},
month = oct,
doi = {10.3389/fbinf.2026.2001675},
url = {https://doi.org/10.3389/fbinf.2026.2001675},
note = {CC BY},
}
RIS
TY - JOUR
TI - SelCysML: a hybrid ML-BLAST pipeline for the automated identification and functional annotation of bacterial selenoproteins
AU - José L. Lavín
AU - José A. Oguiza
PY - 2026
DA - 2026/10/26
JO - Frontiers in Bioinformatics
DO - 10.3389/fbinf.2026.2001675
UR - https://doi.org/10.3389/fbinf.2026.2001675
AB - Selenoproteins are characterized by the incorporation of selenocysteine (Sec) via UGA codon recoding, play vital roles in redox homeostasis and enzymatic catalysis. Despite this biological significance, their computational identification remains challenging, and recent assessments confirm that public databases suffer from widespread annotation errors, often misclassifying these proteins. Current predictive tools frequently operate as “black boxes” with limited interpretability or lack integrated functional validation layers, which contributes to high false-positive rates and persistent ambiguities in the annotation of the selenoproteome. We present SelCysML, an accessible and highly accurate computational pipeline that integrates an ensemble of machine learning (ML) models with a parallel functional validation module using BLAST. By utilizing curated physicochemical descriptors—Amino Acid Composition (AAC) and Composition, Transition, Distribution (CTDC)—the model achieves a robust performance, with an AUC-ROC of 0.968 and a Matthews Correlation Coefficient (MCC) of 0.787. SelCysML not only classifies protein sequences but also provides functional evidence through sequence similarity, effectively bridging the gap between automated prediction and experimental biological insight. SelCysML provides a user-friendly, interpretable, and precise solution for the identification of bacterial selenoproteins in proteomic datasets, serving as a reliable tool for both bioinformaticians and experimental biologists.
SN - 2673-7647
ER -