Deep Forest for Antitubercular Peptides

Overview

Developed a machine learning framework using a deep forest architecture to predict antitubercular peptides, addressing the global challenge of antibiotic resistance.

Key highlights

• Curated and cleaned two benchmark peptide datasets for model training and validation.
• Engineered complex features, including graphical, evolutionary, and binary-profile representations.
• Validated state-of-the-art performance via ablation studies, achieving an F1-score of 94.2% (a 1.1% improvement over prior models).