CDpred

CDpred predicts celiac disease-associated peptides and motifs to identify epitopes in protein and peptide sequences that bind HLA-DQ2 and HLA-DQ8 and may trigger immune responses in celiac disease.

Key Features:

• Epitope Prediction: Predicts peptides associated with celiac disease that interact with HLA-DQ2 and HLA-DQ8 alleles.
• Dataset Utilization: Uses a dataset of experimentally validated celiac disease-associated and non-associated peptides for training, testing, and evaluation.
• Positional Analysis: Performs positional analysis to identify significant amino acid residues within peptides and to assess HLA allele frequency.
• Amino Acid Composition Computation: Computes amino acid composition features from peptides for development of machine learning models.
• Motif-Based Approach: Identifies celiac-associated motifs such as QPF, QPQ, and PYP and evaluates the prevalence of proline (P) and glutamine (Q).
• Machine Learning Models: Trains machine learning models on peptide composition features, reporting a maximum AUROC of 0.99.
• Ensemble Methodology: Combines motif-based approaches with machine learning models in an ensemble that achieved 100% accuracy on independent datasets for predicting CD-associated motifs.

Scientific Applications:

• Peptide screening and design: Screen and design protein or peptide sequences for the presence or absence of celiac disease-associated epitopes.
• Epitope mapping in food and therapeutics: Identify potential epitopes in dietary proteins and therapeutic agents that may bind HLA-DQ2/DQ8 and trigger celiac disease.

Methodology:

Uses experimentally validated peptide datasets for training and testing, conducts positional residue analysis and HLA allele frequency assessment, computes amino acid composition features, identifies motifs (e.g., QPF, QPQ, PYP), trains machine learning models on composition features, and combines motif-based and machine learning approaches in an ensemble evaluated by AUROC and independent dataset accuracy.