MLACP 2.0

MLACP 2.0 predicts anticancer peptides (ACPs) from peptide sequence information using machine learning to prioritize candidate ACPs for anticancer research.

Key Features:

• Enhanced Robustness: Improved robustness relative to the original MLACP, increasing the reliability of ACP predictions.
• Comprehensive Datasets: Uses large non-redundant training and independent datasets curated specifically for ACP research.
• Advanced Feature Encodings: Explores a wide range of feature encodings to represent peptide sequence properties.
• Multi-Classifier Approach: Employs seven conventional classifiers to develop models across different encodings.
• Convolutional Neural Network (CNN) Integration: Selects best-performing encoding-based models per classifier, concatenates their predicted scores, and refines the concatenated scores using a convolutional neural network.
• Superior Performance: Demonstrates improved performance versus recent ACP prediction tools, CNN-based embedding models, and conventional single models in cross-validation and independent evaluations.

Scientific Applications:

• Novel ACP discovery: Prioritizes sequence-derived candidate anticancer peptides for experimental validation and development.
• Experimental design support: Guides hypothesis-driven experiment planning by providing ranked predictions from sequence data.

Methodology:

Constructs large non-redundant training and independent datasets, explores diverse feature encodings, trains seven conventional classifiers on multiple encodings, selects best-performing encoding-based models per classifier, concatenates predicted scores from those models, refines the concatenated scores with a convolutional neural network, and evaluates performance by cross-validation and independent dataset comparisons against recent ACP tools, CNN-based embedding models, and conventional single models.