anticp2

anticp2 predicts anticancer peptides (ACPs) and facilitates sequence-based design of ACPs by applying machine learning to peptide composition for therapeutic and research applications.

Key Features:

• Machine Learning Models: ETree classifiers trained to discriminate ACPs using sequence-derived features.
• Datasets: Two labeled datasets (main and alternate) comprising peptide sequences annotated with anticancer properties.
• Input Features: Dipeptide composition and amino acid composition used as primary feature sets.
• Performance Metrics: On the main dataset an ETree model based on dipeptide composition achieved MCC 0.51 and AUROC 0.83, while on the alternate dataset an amino acid composition-based ETree achieved MCC 0.80 and AUROC 0.97.
• Validation Techniques: Five-fold cross-validation during model development and evaluation on independent validation datasets.
• Sequence-level Insights: Identified residue preferences (A, F, K at N-terminus; L, K at C-terminus) and recurring motifs such as LAKLA, AKLAK, FAKL, and LAKL.

Scientific Applications:

• Peptide Design: Predicts potential anticancer activity of peptide sequences to guide rational design of novel ACPs.
• Therapeutic Development: Supports development of peptide-based therapeutics in oncology and strategies for personalized medicine.
• Research Insights: Provides residue preference and motif analyses to inform sequence–activity relationships and structure-function studies.

Methodology:

Residue composition analysis, motif identification, and machine learning implementation using ETree classifiers trained on main and alternate datasets with dipeptide and amino acid composition features, evaluated by five-fold cross-validation and independent validation.