Accurate informatic modeling of tooth enamel pellicle interactions by training substitution matrices with Mat4Pep

Motivation Protein-hydroxyapatite interactions govern the development and homeostasis of teeth and bone. Characterization would enable design of peptides to regenerate mineralized tissues and control attachments such as ligaments and dental plaque. Progress has been limited because no available methods produce robust data for assessing phase interfaces. Results We show that tooth enamel pellicle peptides contain subtle sequence similarities that encode hydroxyapatite binding mechanisms, by segregating pellicle peptides from control sequences using our previously developed substitution matrix-based peptide comparison protocol (Oren et al., 2007), with improvements. Sampling diverse matrices, adding biological control sequences, and optimizing matrix refinement algorithms improves discrimination from 0.81 to 0.99 AUC in leave-one-out experiments. Other contemporary methods fail on this problem. We find hydroxyapatite interaction sequence patterns by applying the resulting selected refined matrix (“pellitrix”) to cluster the peptides and build subgroup alignments. We identify putative hydroxyapatite maturation domains by application to enamel biomineralization proteins and prioritize putative novel pellicle peptides identified by In stageTip (iST) mass spectrometry. The sequence comparison protocol outperforms other contemporary options for this small and heterogeneous group, and is generalized for application to any group of peptides. Availability Software to apply this protocol is freely available at github.com/JeremyHorst/Mat4Pep and compbio.org/protinfo/ Mat4Pep. Contact [email protected] , [email protected] . Supplementary information Available at Bioinformatics online.