Classification of Membrane Proteins by Types of Transmembrane Helices Using SOSUI System

There are many amino acid sequences in proteomes which are not homologous to any other sequences. Therefore, methods to classify proteins independent of the sequence homology are strongly required for computational analysis of proteomes. Proteins may be divided into two categories: soluble and membrane proteins. Since membrane proteins are characterized by the existence of long hydrophobic transmembrane helices, the classification of amino acid sequences into two types of proteins, soluble and membrane proteins, is possible with considerably high accuracy. The classification of all amino acid sequences in several proteomes was reported recently, leading to the conclusion that the fraction of membrane proteins is about 30% [1]. However, those methods provided only the information about the number of transmembrane helices. We have previously proposed a new method (SOSUI) to classify amino acid sequences by two types of transmembrane helices, primary and secondary transmembrane segments. In this work, we have analyzed four kinds of dataset including two set of all amino acid sequences from proteomes, using the SOSUI system [2] which not only predicts transmembrane helix regions but also classifies them by the strength of interaction with lipid membranes.