SAPPTree: Identification of an S-Acylation Motif Drives a Novel S-Acylation Prediction Program

S-acylation, the reversible addition of fatty acids to proteins, has emerged as an abundant post-translational modification that drives protein localization and function. With no known consensus sequence, current prediction programs rely on machine learning algorithms that use short peptide sequences and large proteomic datasets. However, current prediction programs often suggest incorrect sites of S-acylation, leading to wasted experimental time and effort following site-directed mutagenesis and low-throughput validation experiments. Using only experimentally confirmed sites of S-acylation, we sought to identify primary sequence, secondary structure, and tertiary structure features common amongst S-acylation sites to aid in developing more robust prediction tools. In doing so, we identified an S-acylation motif including a cysteine cluster flanked by a hydrophobic stretch, and a positively charged polybasic region found within a helical stretch. These features were combined with known or AlphaFold-predicted structures and additional features including residue depth and solvent accessibility into a random forest model to generate a new and more accurate S-acylation prediction program (SAPP), named SAPPTree. All the processed datasets and complete model training pipeline are available at https://github.com/neurdyphagy-lab/palm-prediction-model, while the webserver is available at http://martintools.sci.uwaterloo.ca/.