Detecting Subtle Sequence Signals: a Gibbs Sampling Strategy for Multiple Alignment

Abstract

A wealth of protein and DNA sequence data is being generated by genome projects and other sequencing efforts. A crucial barrier to deciphering these sequences and understanding the relations among them is the difficulty of detecting subtle local residue patterns common to multiple sequences. Such patterns frequently reflect similar molecular structures and biological properties. A mathematical definition of this "local multiple alignment" problem suitable for full computer automation has been used to develop a new and sensitive algorithm, based on the statistical method of iterative sampling. This algorithm finds an optimized local alignment model for N sequences in N-linear time, requiring only seconds on current workstations, and allows the simultaneous detection and optimization of multiple patterns and pattern repeats. The method is illustrated as applied to helix-turn-helix proteins, lipocalins, and prenyltransferases.

Gibbs motif sampling: Detection of bacterial outer membrane protein repeats

“…How does it differ from other methods (including the site sampler) and under what circumstances is it to be preferred? Lawrence et al (1993) have compared Gibbs sampling with several other motif methods and this need not be reiterated here. More recently, however, some closely related methods that utilize HMM for multiple sequence alignment have been described (Baldi et al, 1994;Krogh et al, 1994).…”

“…Only more recently, however, have efficient methods been developed that can detect subtle similarities common to large sets of distantly related or (possibly) evolutionarily unrelated sequences (Lawrence et al, 1993;Neuwald & Green, 1994). The development of these methods has been motivated by the current rapid increase in sequence data because relatively large sets (containing, for example, more than 15 sequences) are needed for weakly conserved patterns to reach statistical significance.…”

“…The development of these methods has been motivated by the current rapid increase in sequence data because relatively large sets (containing, for example, more than 15 sequences) are needed for weakly conserved patterns to reach statistical significance. Lawrence et al (1993) describe a Gibbs sampling strategy for detecting conserved patterns in multiple sequences that is a stochastic analog of earlier expectation-maximization methods (Lawrence & Reilly, 1990;Cardon & Stormo, 1992) and that is closely related to (EM-based) hidden Markov model multiple sequence alignment methods (Baldi et al, 1994;Krogh et al, 1994), which, unlike the Gibbs sampler, permit gaps anywhere in the sequences. This Gibbs sampler (which is referred to here as the site sampler) addresses the problem of finding motifs when the number of occurrences of each motif in each sequence is assumed.…”

Finding flexible patterns in unaligned protein sequences

“…On the other hand, these methods, as well as Roytberg (1992) and Saqi and Sternberg (1994), are likely to find patterns that Pratt will not find. The Gibbs sampler-based method (Lawrence et al, 1993) is superior to Pratt at aligning weakly conserved regions lacking strongly conserved positions but is unable to find patterns having variable length spacing between conserved positions.…”

P100, a transcriptional coactivator, is a human homologue of staphylococcal nuclease

“…Selfcomparisons of the pl00 sequence using dot plots (Thompson et al, 1994a) and REPRO, an algorithm that detects distant repeat sequences (Heringa & Argos, 1993), indicated the presence of four repeats each containing approximately 150 residues. Analysis using MACAW (Schuler et al, 1991) indicated that the four repeats in human pl00 are homologues, as multiple alignments generated by Gibbs sampling (Lawrence et al, 1993) revealed the probabilities (p-values) of four separate blocks aligning by chance of 3.4 X IO"', 2.3 X 4.2 X and 9.7 X IO" (here a maximal searchspace N was chosen as N = 8854; human pl00 contains 885 residues). Similar results (not shown) were obtained for a second pl00 sequence (FIOg7.2), known as a result of the Caenorhabdiris elegans genome project (Wilson et al, 1994).…”