B-LacFamPred

B-LacFamPred predicts and classifies β-lactamases from sequence data to assign Ambler class, subclass, and family for studies of β-lactamase diversity and antibiotic resistance.

Key Features:

• Classification System: Categorizes β-lactamases according to Ambler's classification into classes A–D with further subdivision into subclasses and 96 families to reflect hydrolytic profiles.
• Machine Learning-Based Approach: Employs a machine-learning in silico method using Hidden Markov Models (HMMs) trained on sequence data for prediction and annotation.
• High Accuracy: HMM classifiers achieved 100% accuracy in all but one instance during leave-one-out cross-validation in validation tests.
• Benchmarking Success: Benchmarked against other β-lactamase family prediction methods, correctly predicted 56 of 60 penicillin-binding proteins (PBPs) and did not misidentify glyoxalase II sequences as β-lactamases.
• Reduced False Positives: Exhibits a lower rate of false-positive predictions in proteome-wide annotation tasks compared to tools such as DeepBL.

Scientific Applications:

• Antibiotic Resistance Research: Enables classification of β-lactamases to support studies of resistance mechanisms.
• Evolutionary Analysis: Supports investigations into the evolutionary dynamics of β-lactamase families and subclasses.
• Antimicrobial Development: Provides family- and subclass-level annotation that can inform development of new antimicrobial strategies.
• Proteome-wide Annotation: Facilitates large-scale proteome annotation with reduced false positives for accurate identification of β-lactamases.

Methodology:

Uses Hidden Markov Models (HMMs) trained on sequence data to predict Ambler class, subclass, and one of 96 β-lactamase families, with validation via leave-one-out cross-validation and benchmarking against other methods (including DeepBL) and test sets such as PBPs and glyoxalase II sequences.