Biomedical Engineering Reference
In-Depth Information
Bayesian Methods
Although not considered mainstream by many researchers in the bioinformatics field, Bayesian
statistical methods can be used to determine pairwise sequence alignment and to estimate the
evolutionary distance between DNA sequences. Bayesian methods involve examining the probabilities
of all possible alignments, gap scores, and substitution matrix values (the prior probabilities) to
assess the probability of an alignment (the posterior probability). Proponents of the Bayesian
approach to sequence alignment cite as advantages over the limitations of dynamic program the
method's ability to fully and exactly describe uncertainty, derive exact significance measures, and
eliminate the need to specify all parameters.
In practice, Bayesian-based tools, such as the Bayes Block Aligner, a workstation-based tool available
from the Center for Bioinformatics at Rensselaer and Wadsworth Center of the New York Department
of Health, performs better than dynamic programming in some cases, and not as well in others. The
Block Aligner manipulates two sequences to find the highest-scoring contiguous regions (blocks),
which are then joined in various combinations to form alignments. Unlike a dynamic programming or
word-based approach, the Bayes Block Aligner, which works with both DNA and protein sequences,
doesn't require the user to specify a particular substitution matrix or gap scoring system. Instead, it
bases the posterior probability distributions of alignments on the number of blocks expected in an
alignment and a range of substitution matrices. A Web-based Bayesian analysis tool, the Bayesian
Algorithm for Local Sequence Alignment (BALSA) is also available from the center. BALSA is
described in the "
Tools
" section later in this chapter.
