dmppred

dmppred predicts, designs, and scans peptides associated with type 1 diabetes mellitus (T1DM) to identify antigenic regions (epitopes) relevant for peptide-based immunotherapy.

Key Features:

• Innovative Prediction Methodology: Combines alignment-based and alignment-free approaches because the analyzed T1DM-associated peptides do not show specific association with HLA alleles.
• Alignment-Based Method: Uses Basic Local Alignment Search Tool (BLAST) similarity/alignment searches that provide a high probability of correct hits but limited coverage.
• Alignment-Free Method: Employs machine learning using dipeptide composition and achieved a maximum AUROC of 0.89.
• Hybrid Approach: Integrates alignment-based and alignment-free predictions, yielding an AUROC of 0.95 and a Matthews correlation coefficient (MCC) of 0.81 on independent datasets.
• Dataset Analysis: Based on a comprehensive analysis of 815 T1DM-associated peptides revealing non-specific association with HLA alleles.

Scientific Applications:

• Antigen-Specific Immunotherapy Development: Supports prediction and design of T1DM-associated peptides for peptide-based immunotherapies.
• Therapeutic Target Identification: Aids identification of antigenic regions/epitopes as potential therapeutic targets.
• Autoimmunity Research: Facilitates studies of autoimmune responses against pancreatic β-cells and design of peptide-based interventions to modulate immune responses.

Methodology:

Computational methods include BLAST-based alignment searches, machine learning classifiers using dipeptide composition, integration of alignment-based and alignment-free predictions into a hybrid model, and analysis of 815 T1DM-associated peptides revealing non-specific HLA allele association.