ACPred-Fuse

ACPred-Fuse predicts anticancer peptides from protein sequences by fusing 29 sequence-based feature descriptors and handcrafted sequential features with machine learning to identify peptides with potential anticancer activity.

Key Features:

• Machine learning predictor: Employs a novel machine learning model to assess whether protein-derived peptides function as anticancer peptides (ACPs).
• Multiview feature integration: Integrates 29 diverse sequence-based feature descriptors that capture class and probabilistic information with handcrafted sequential features.
• Feature representation learning: Learns embedded class and probabilistic representations from ACP-related features to improve discriminatory power.
• Multiview fusion for optimized representation: Fuses complementary feature types to enhance the representational capacity of input data for model training.
• Benchmarking performance: Demonstrates superior predictive precision and reliability relative to existing ACP predictors in comparative analyses.

Scientific Applications:

• In silico ACP screening: Enables high-throughput computational identification of candidate anticancer peptides from protein sequences.
• Candidate prioritization for experimental validation: Ranks predicted peptides to guide selection for biochemical and cellular assays.
• Support for anticancer peptide discovery: Provides computational evidence to inform peptide-based drug discovery and therapeutic development efforts.

Methodology:

Feature representation learning to capture embedded class and probabilistic information in ACPs; multiview feature fusion combining 29 sequence-based descriptors with handcrafted sequential features; and training and optimization of a machine-learning prediction model using the fused multiview features.