Protein Digestion Simulator

Protein Digestion Simulator predicts peptide reversed-phase liquid chromatography (RP-LC) retention times and simulates enzymatic digestion of protein or peptide sequences for proteomics analyses.

Key Features:

• Input/Output Formats: Reads protein or peptide sequences in FASTA format or delimited text files and writes results to tab-delimited files.
• Enzymatic Digestion: Performs sequence digestion using various enzymatic rules, including trypsin and partial trypsin protocols.
• ANN-Based Retention Time Prediction: Predicts peptide retention times in RP-LC using an artificial neural network (ANN) model.
• Peptide Descriptors: Incorporates peptide length, sequence, hydrophobicity, hydrophobic moment, nearest-neighbor amino acids, and predicted secondary structures (helix, sheet, coil) as model inputs.
• Model Architecture: Implements an ANN with 1,052 input nodes, 24 hidden nodes, and a single output node.
• Training Dataset: Trained on approximately 345,000 nonredundant peptides derived from over 12,059 LC-MS/MS analyses across more than 20 organisms.
• Validation Performance: Evaluated on 1,303 confidently identified peptides excluded from training, achieving an average elution time precision of approximately 1.5%.
• Isomer Discrimination: Distinguishes among isomeric peptides by incorporating peptide sequence information into predictions.

Scientific Applications:

• Proteomics Studies: Supports proteomic experiments requiring accurate peptide retention time prediction.
• Chromatography Optimization: Aids optimization of reversed-phase LC conditions based on predicted peptide elution behavior.
• Mass Spectrometry Interpretation: Enhances interpretation of LC-MS/MS data by providing predicted retention times for peptides.
• Protein Identification and Characterization: Contributes retention time evidence to improve peptide and protein identification and characterization workflows.

Methodology:

Parse FASTA or delimited text inputs, perform enzymatic digestion using specified rules (including trypsin and partial trypsin), compute peptide descriptors (length, sequence, hydrophobicity, hydrophobic moment, nearest-neighbor residues, predicted helix/sheet/coil), input descriptors to an ANN with 1,052 input nodes, 24 hidden nodes, and one output node, train on ~345,000 nonredundant peptides from >12,059 LC-MS/MS analyses across >20 organisms, validate on 1,303 withheld peptides (≈1.5% average elution time precision), and output results as tab-delimited files.