SVMyr

SVMyr predicts co- and post-translational myristoylation sites in proteins by using support vector machines trained on composition and physicochemical features of N-terminal octapeptides to identify N-myristoyltransferase (NMT) substrates.

Key Features:

• Algorithm: Support Vector Machines (SVM) model trained to classify myristoylation potential of N-terminal octapeptides.
• Feature encoding: Composition and physicochemical properties of octapeptides encoding interactions with N-myristoyltransferases (NMTs).
• Co- and post-translational prediction: Predicts both co-translational myristoylation at the exposed N-terminal glycine and post-translational myristoylation following proteolytic cleavage.
• Caspase cleavage integration: Incorporates detection of upstream caspase cleavage sites by searching for regular motifs documented in the literature to enable post-translational site prediction.
• Cross-validation performance: Reported cross-validation metrics include AUC = 0.92 and MCC = 0.61.
• Independent benchmark: Validated on an independent dataset (88 medium/high confidence co-translational sites and 528 negatives) yielding AUC = 0.91 and MCC = 0.58.
• UniProt validation: Recovers 96% of electronically annotated myristoylated proteins in UniProt (31,048 entries).
• Proteome-wide identification: Identifies potential myristoylomes across eight proteomes and reports new putative substrates for NMTs.

Scientific Applications:

• Site prediction: Identification of putative co- and post-translational myristoylation sites in protein sequences.
• Functional annotation: Supporting annotation of protein function related to membrane targeting, signal transduction, and protein–protein interactions via myristoylation status.
• NMT substrate discovery: Discovery of putative N-myristoyltransferase substrates and expansion of known myristoylomes across multiple proteomes.
• Proteolysis-linked modification studies: Studying the interplay between caspase-mediated cleavage and subsequent post-translational myristoylation.

Methodology:

SVM models trained on composition and physicochemical features of N-terminal octapeptides; cross-validation reporting AUC and MCC; benchmarking on an independent dataset of 88 positives and 528 negatives; motif-based search for upstream caspase cleavage sites to predict post-translational myristoylation; validation against UniProt electronic annotations and proteome-wide scanning across eight proteomes.