LSTM-assisted machine learning framework for low-data prioritization of CepI-targeted candidates: a focus on Burkholderia cenocepacia

Introduction Antimicrobial resistance poses a substantial threat to global healthcare particularly in immunocompromised individuals. Burkholderia cenocepacia is a clinically important opportunistic bacterium responsible for persistent respiratory infections in cystic fibrosis patients. Its intrinsic antimicrobial resistance and ability to form biofilms make conventional treatment ineffective. Therefore, targeting virulence-associated quorum sensing pathway represents an alternative anti-virulence strategy to reduce pathogenicity without directly imposing bactericidal pressure. Importantly, CepI , an autoinducer synthase involved in acyl-homoserine lactone biosynthesis, serves as a promising molecular target for disrupting quorum sensing mediated pathogenicity. Methods The present study combines randomized-SMILES augmentation, LSTM-based molecular generation, machine-learning-aided drug repurposing and structure-based validation to identify potential CepI -targeting candidates. Model robustness was evaluated using scaffold-aware, y-randomization, provenance-control and cross-provenance. The optimized model was applied to the DrugBank library, followed by applicability-domain assessment using Mahalanobis distance, k-nearest-neighbor analysis and principal component analysis. Prioritized candidates were further evaluated by docking, MM-GBSA, ADMET analysis and triplicate molecular dynamics simulations. Results and discussion Among the evaluated classifiers, the Extra Trees model showed the best 10-fold cross-validation performance, with a ROC-AUC of 0.992 ± 0.025. In contrast, 100-iteration Y-randomization produced a mean ROC-AUC of 0.481 ± 0.084 (maximum = 0.559), compared with 0.99 for the original model (p = 0.0099). The near-chance performance under randomized labels provided evidence against a simple chance-label explanation. Provenance-control analyses indicated that the classification signal was not explained solely by a simple literature-versus-LSTM source label, although residual provenance and chemical-space bias cannot be completely excluded. Integration of the activity-probability threshold and applicability-domain criteria narrowed the DrugBank screening set to 115 compounds for structure-based analysis. Among these, DB12364 (betrixaban) was prioritized based on its predicted activity, applicability-domain assessment, docking, MM-GBSA and molecular dynamics results, and is proposed as a computationally prioritized candidate for experimental evaluation as a potential CepI -targeting compound against B. cenocepacia .

Authors

Institutions

Publication Details

Journal
Frontiers in Cellular and Infection Microbiology
Published
2026-11-18
DOI
https://doi.org/10.3389/fcimb.2026.1934561
Primary Topic
Computational Drug Discovery Methods
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

LSTM-assisted machine learning framework for low-data prioritization of CepI-targeted candidates: a focus on Burkholderia cenocepacia

Vijayavidhya Magizhvannan, Shanthi Veerappapillai
3 links
Frontiers in Cellular and Infection Microbiology
Computational Drug Discovery Methods
article

LSTM-assisted machine learning framework for low-data prioritization of CepI-targeted candidates: a focus on Burkholderia cenocepacia

Vijayavidhya Magizhvannan, Shanthi Veerappapillai
article en

Abstract

Introduction Antimicrobial resistance poses a substantial threat to global healthcare particularly in immunocompromised individuals. Burkholderia cenocepacia is a clinically important opportunistic bacterium responsible for persistent respiratory infections in cystic fibrosis patients. Its intrinsic antimicrobial resistance and ability to form biofilms make conventional treatment ineffective. Therefore, targeting virulence-associated quorum sensing pathway represents an alternative anti-virulence strategy to reduce pathogenicity without directly imposing bactericidal pressure. Importantly, CepI , an autoinducer synthase involved in acyl-homoserine lactone biosynthesis, serves as a promising molecular target for disrupting quorum sensing mediated pathogenicity. Methods The present study combines randomized-SMILES augmentation, LSTM-based molecular generation, machine-learning-aided drug repurposing and structure-based validation to identify potential CepI -targeting candidates. Model robustness was evaluated using scaffold-aware, y-randomization, provenance-control and cross-provenance. The optimized model was applied to the DrugBank library, followed by applicability-domain assessment using Mahalanobis distance, k-nearest-neighbor analysis and principal component analysis. Prioritized candidates were further evaluated by docking, MM-GBSA, ADMET analysis and triplicate molecular dynamics simulations. Results and discussion Among the evaluated classifiers, the Extra Trees model showed the best 10-fold cross-validation performance, with a ROC-AUC of 0.992 ± 0.025. In contrast, 100-iteration Y-randomization produced a mean ROC-AUC of 0.481 ± 0.084 (maximum = 0.559), compared with 0.99 for the original model (p = 0.0099). The near-chance performance under randomized labels provided evidence against a simple chance-label explanation. Provenance-control analyses indicated that the classification signal was not explained solely by a simple literature-versus-LSTM source label, although residual provenance and chemical-space bias cannot be completely excluded. Integration of the activity-probability threshold and applicability-domain criteria narrowed the DrugBank screening set to 115 compounds for structure-based analysis. Among these, DB12364 (betrixaban) was prioritized based on its predicted activity, applicability-domain assessment, docking, MM-GBSA and molecular dynamics results, and is proposed as a computationally prioritized candidate for experimental evaluation as a potential CepI -targeting compound against B. cenocepacia .

Frontiers in Cellular and Infection MicrobiologyVol. 16
Vellore Institute of Technology University (IN)
Good health and well-being
Computational Drug Discovery Methods
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