ACP-MLC

ACP-MLC predicts anticancer peptides (ACPs) and classifies their tissue-type targets using a two-level machine-learning framework for peptide identification and multi-label tissue targeting.

Key Features:

• Two-Level Prediction Engine: A first-level random forest model identifies ACPs and a second-level binary relevance algorithm performs multi-label classification of tissue-type targets.
• Performance Metrics: First-level prediction achieved an AUC of 0.888 on independent test datasets; second-level prediction reported hamming loss 0.157, subset accuracy 0.577, F1-score_macro 0.802, and F1-score_micro 0.826 on independent test sets.
• Comparative Advantage: Systematic comparisons indicate performance superior to existing binary classifiers and multi-label learning classifiers for ACP prediction.
• Feature Interpretation: SHAP (SHapley Additive exPlanations) is used to interpret and rank feature contributions to predictions.

Scientific Applications:

• Mechanistic Insights: Identification and feature interpretation of ACPs can inform hypotheses about peptide mechanisms of action.
• Therapeutic Development: Predictions of ACP identity and tissue-specific targets support selection of peptide candidates for anticancer therapy development.
• Research Efficiency: Automated ACP identification and multi-label tissue prediction streamline candidate prioritization in peptide research.

Methodology:

First-level random forest classification for ACP identification; second-level binary relevance multi-label classification for tissue-type prediction; SHAP for feature interpretation; evaluation on independent test datasets and systematic comparison with existing binary and multi-label classifiers.