Protein Structure and Function: Application of Bioinformatics Methods - John Rigden 2014

Prediction of Membrane Protein Structure
Multiple Sequence Alignments

Multiple sequence alignments play a crucial role in predicting The Structure of TM Proteins. Homologous sequences retrieved from Databases can be used to generate sequence profiles, which significantly enhance the accuracy of TM topology prediction (Kall et al. 2005; Jones 2007), while identified templates can be utilized for Homology modeling.

When applying traditional pairwise alignment Methods, sequences suitable for the protein under investigation are selected based on scoring function values. These values are derived using amino acid Substitution Matrices, such as PAM (Dayhoff et al. 1978) or BLOSUM (Henikoff and Henikoff 1992). Because such matrices were originally developed for aligning Globular proteins—and globular and TM proteins differ significantly in their Amino Acid Composition, Hydrophobicity, and The Nature of conserved regions (Jones et al. 1994a)—they are fundamentally unsuited for aligning TM proteins. Consequently, specialized substitution matrices tailored to the unique properties of transmembrane proteins have been developed. For instance, the JJT TM matrix (Jones et al. 1994b) was designed accounting for the fact that polar residues in transmembrane proteins exhibit high conservation, whereas hydrophobic residues are readily interchangeable. The SLIM matrix (Muller et al. 2001) delivers high accuracy in identifying distant homologs within a curated GPCR dataset. The PHAT matrix (Ng et al. 2000) outperforms JJT in performance metrics, particularly in database searching. Nevertheless, to date, no independent studies have evaluated TM substitution matrices using a common benchmark dataset.

Several novel methods offer higher alignment quality for TM proteins. In STAM (Sharif and Gut 2004), the penalty imposed for insertions/deletions within TM segments is higher than the corresponding penalty for insertions/deletions in loop regions, and the method also employs different substitution matrices. As a result, homology models built using this approach exhibit greater accuracy. PRALINETM (Pirovano et al. 2008) combines state-of-the-art sequence prediction techniques with membrane-specific substitution matrices, having been shown to outperform standard multiple alignment methods such as ClustalW (Higgins et al. 1994) or Muscle (Edgar 2004). The study utilized the BaliBASE benchmark dataset (Bahr et al. 2001) to evaluate the quality of transmembrane alignments. Recent modifications introduced to BLAST and PSI-BLAST (Altschul and Koonin 1998) to account for the COMPOSITION OF THE query sequence should theoretically improve TM protein search performance (Altschul et al. 2005), although independent studies have likewise not been conducted in this case. The advanced T-Coffee alignment method (Notredame et al. 2000), despite using a single general scoring matrix, delivers high-quality results in cases of high sequence identity, as demonstrated by tests on a benchmark set of homologous membrane protein structures. In HMAP (Tang et al. 2003), a substantial improvement in alignment quality is achieved through profile analysis and comparison incorporating structural information.