Abstract
We describe an improved artificial neural network (ANN)-based method for predicting peptide retention times in reversed-phase liquid chromatography. In addition to the peptide amino acid composition, this study investigated several other peptide descriptors to improve the predictive capability, such as peptide length, sequence, hydrophobicity and hydrophobic moment, and nearest-neighbor amino acid, as well as peptide predicted structural configurations (i.e., helix, sheet, coil). An ANN architecture that consisted of 1052 input nodes, 24 hidden nodes, and 1 output node was used to fully consider the amino acid residue sequence in each peptide. The network was trained using approximately 345,000 nonredundant peptides identified from a total of 12,059 LC-MS/MS analyses of more than 20 different organisms, and the predictive capability of the model was tested using 1303 confidently identified peptides that were not included in the training set. The model demonstrated an average elution time precision of approximately 1.5% and was able to distinguish among isomeric peptides based upon the inclusion of peptide sequence information. The prediction power represents a significant improvement over our earlier report (Petritis, K.; Kangas, L. J.; Ferguson, P. L.; Anderson, G. A.; Pasa-Tolic, L.; Lipton, M. S.; Auberry, K. J.; Strittmatter, E. F.; Shen, Y.; Zhao, R.; Smith, R. D. Anal. Chem. 2003, 75, 1039-1048) and other previously reported models.