Computing Prediction and Functional Analysis of Prokaryotic Propionylation

Identification and
systematic analysis of candidates for protein
propionylation are crucial steps for understanding its molecular mechanisms
and biological functions. Although several proteome-scale methods
have been performed to delineate potential propionylated proteins,
the majority of lysine-propionylated substrates and their role in
pathological physiology still remain largely unknown. By gathering
various databases and literatures, experimental prokaryotic propionylation
data were collated to be trained in a support vector machine with
various features via a three-step feature selection method. A novel
online tool for seeking potential lysine-propionylated sites (PropSeek)
(http://bioinfo.ncu.edu.cn/PropSeek.aspx) was built. Independent test results of leave-one-out and n-fold cross-validation were similar to each other, showing
that PropSeek is a stable and robust predictor with satisfying performance.
Meanwhile, analyses of Gene Ontology, Kyoto Encyclopedia of Genes
and Genomes pathways, and protein–protein interactions implied
a potential role of prokaryotic propionylation in protein synthesis
and metabolism.