PASTA sequence composition is a predictive tool for protein class identification

PASTA domains are small modules expressed in bacteria and found in one or multiple copies at the C-terminal end of several penicillin binding proteins (PBPs) and Ser/Thr protein kinases (STPKs) and represent potential targets for a new class of antibiotics. PASTA domains are currently annotated as sensor domains, as they are thought to activate their cognate proteins in response to binding to opportune ligands. However, recent studies have shown that PASTA domains linked to proteins of different classes, STPKs or PBPs, do not share the same binding abilities. Despite this, there is currently no way to distinguish between PASTA domains from the two classes, since all of them share the same fold, independent of the class they belong to. To identify a predictive tool of class identification, we here analyse a pool of parameters, including amino acid compositions and total charges of PASTA domains either linked to PBPs or to STPKs. We screened sequences from Actinobacteria, Firmicutes and Bacteroidetes. The first two phyla include some of the most dangerous micro-organisms for human health such as Mycobacterium tuberculosis and Staphylococcus aureus. Based on this analysis, our study proposes a predictive method to assign PASTA domains with unknown origin to their corresponding enzyme class, based solely on sequence information.