Alignment of protein sequences by their profiles

Abstract

The accuracy of an alignment between two protein sequences can be improved by including other detectably related sequences in the comparison. We optimize and benchmark such an approach that relies on aligning two multiple sequence alignments, each one including one of the two protein sequences. Thirteen different protocols for creating and comparing profiles corresponding to the multiple sequence alignments are implemented in the SALIGN command of MODELLER. A test set of 200 pairwise, structure-based alignments with sequence identities below 40% is used to benchmark the 13 protocols as well as a number of previously described sequence alignment methods, including heuristic pairwise sequence alignment by BLAST, pairwise sequence alignment by global dynamic programming with an affine gap penalty function by the ALIGN command of MODELLER, sequence-profile alignment by PSI-BLAST, Hidden Markov Model methods implemented in SAM and LOBSTER, pairwise sequence alignment relying on predicted local structure by SEA, and multiple sequence alignment by CLUSTALW and COMPASS. The alignment accuracies of the best new protocols were significantly better than those of the other tested methods. For example, the fraction of the correctly aligned residues relative to the structure-based alignment by the best protocol is 56%, which can be compared with the accuracies of 26%, 42%, 43%, 48%, 50%, 49%, 43%, and 43% for the other methods, respectively. The new method is currently applied to large-scale comparative protein structure modeling of all known sequences.

Detecting remotely related proteins by their interactions and sequence similarity

“…It can be done by establishing (i) a sequence and͞or structure similarity to another characterized protein and (ii) a functional link to another characterized protein. The first group of methods includes sequence matching and threading (9,11). The second group includes both experimental and computational methods, such as clustering by physical interactions (26), mRNA array expression profiles (38), analysis of gene fusion (30), phylogenetic profiles (39), and genomic association of genes (40).…”

“…To minimize sequence matching problems, additional methods, such as profileprofile searches (9), hidden Markov models (10), threading (44), and intermediate sequence search (41) can be used. With the interactions, false positives rate and coverage can be improved by probabilistic methods that rely on multiple sources of information about protein interactions (45) and by performing more experi- (53), with a C ␣ rms deviation of 1.7 Å and 15% sequence identity.…”

“…The former group includes the sequence profile-based methods (8,9) and hidden Markov models (10) that construct a multiple sequence alignment of the close homologs of the query, followed by scanning the corresponding profile against a database of sequences. The latter group includes sequence-structure threading methods that can sometimes reveal more distant relationships than purely sequence-based methods (11).…”

“…The corrected figure and its legend appear below. In addition, on page 14423, the opening sentence of the first paragraph in the second column, which reads ''Total GC-skew analysis (9,16) did not identify any clear origin of replication (OriC), but cumulative GC-skew at the third codon position using ORILOC (17) showed reversal in 6 regions (Fig. 1B),'' should be replaced by the following: ''Cumulative GCskew analysis at the third codon position using ORILOC (17) showed reversal in two regions (Fig.…”

Protein Structure Prediction using String Kernels

“…First, as was the case with a number of studies on the accuracy of protein sequence alignment [38,57,36], the proper use of sequence profiles leads to dramatic improvements in the overall ability to detect remote homologs and identify proteins that share the same structural fold. Second, kernel functions that are constructed by directly taking into account the similarity between the various protein sequences tend to outperform schemes that are based on a feature-space representation (where each dimension of the space is constructed as one of -possibilities in a -residue long subsequence or using structural motifs (Isites) in the case of SVM-HMMSTR).…”