BIOINFORMATICS DISCOVERY NOTE Vol. 16 no. 9 2000 Pages 839–842 Profiles from structure based sequence alignment of porins can identify β stranded integral membrane proteins T. V. Gnanasekaran, Suraj Peri, A. Arockiasamy and S. Krishnaswamy ∗ Bioinformatics Centre, School of Biotechnology, Madurai Kamaraj University, Madurai, India Received on October 13, 1999; revised on February 18, 2000; accepted on April 26, 2000 Supplementary information: http://gene.tn.nic.in/sk/ profiles/ Contact: krishna@mrna.tn.nic.in Introduction Integral membrane protein structures constitute less than 2% of protein structures in the Protein Data Bank (PDB) (Bernstein et al., 1977). The known membrane protein structures can be divided into two structural classes: α helical and β stranded proteins. The β stranded membrane proteins form β barrel structures of which porins are type members. Porins are general diffusion, pore-forming proteins which are present in the outer membrane of Gram-negative bacteria. They exist as homotrimers on the outer membrane and form the majority of outer membrane proteins. The basic structural motif of the porins, the β - barrel that forms the transmembrane core, is formed of 16 β strands in general-diffusion porins and 18 β strands in sugar-specific porins (Schulz, 1996). Earlier Gibbs sampler techniques have been used to detect motifs in bacterial membrane proteins (Neuwald et al., 1995) and to identify transmembrane β -strand regions in mitochondrial pore proteins (Mannella et al., 1996). We have used the profiles made from structurally conserved regions (SCR) of porins as a potential tool to identify or discriminate β stranded integral membrane proteins. Systems and methods The calculations were carried out on an alpha 2000 workstation running OSF/1 DEC Unix v3.2 and Silicon Graphics O2 running IRIX 6.1. The superposition of structures was carried out using BIOSYM software (Biosym Technologies Inc., USA). Profiles were created based on the structure based multiple sequence alignment and analyzed using GCG software (Wisconsin Package Version 9.0, Genetics Computer Group, Wisconsin, ∗ To whom correspondence should be addressed. 0% 20% 40% 60% 80% 100% 1 3 5 7 9 11 13 15 SCR profiles u/nm alpha beta Fig. 1. The classification of the first 20 entries identified by PROFILESEARCH using the profiles (Pscr 1–16) are given as a stacked percentage plot. Three classes of proteins are noted: β stranded membrane proteins, α helical membrane proteins and unknown/non-membrane proteins. About 10–20% false positives were obtained. USA). Analysis was carried out using a non-redundant predominantly membrane protein database (pmpDB nr ) created from the OWL database (Bleasby et al., 1994) (http://bmbsgi11.leeds.ac.uk/bmb5dp/owl.html) using BLAST (Altschul et al., 1990). Throughout, Perl (Wall et al., 1996) scripts and modules were used for sorting, mapping and linking purposes. Algorithm Selected porin structures were superimposed and the structurally conserved transmembrane β strand regions (SCRs) were identified. The structure based sequence alignment was used to generate profiles (Pscr) for each of the SCRs. These profiles were used to search pmpDB nr . The profiles were compared by their ability to discrimi- nate in the pmpDB nr : (a) identified β stranded membrane proteins in the top 20 hits (b) structurally characterized β stranded integral membrane proteins other than those used for the profile construction and (c) more β stranded than α helical membrane proteins. The profiles were then ranked based on their performance. c  Oxford University Press 2000 839