TportHMM: Predicting the substrate class of transmembrane transport proteins using profile Hidden Markov Models

Transporters make up a large proportion of proteins in a cell, and play important roles in metabolism, regulation, and signal transduction by mediating movement of compounds across membranes. There is a need for tools that predict the substrates which are transported at the level of substrate class and the level of specific substrate. Our work develops a predictor, TportHMM, using profile Hidden Markov Model (HMM). We explore the role of multiple sequence alignment (MSA) algorithms to utilise evolutionary information, specificity-determining site (SDS) algorithms to highlight positional information, and a profile Hidden Markov Model (HMM) classifier to utilise sequence information. We study the impact of different MSA algorithms (ClustalW, Clustal Omega, MAFFT, MUSCLE, AQUA, T-Coffee and TM-Coffee), and different SDS algorithms (Speer Server, GroupSim, Xdet and TCS). We compare these approaches with the state-of-the-art, TrSSP and TranCEP.