This article is part of the supplement: Selected articles from the Eighth Asia-Pacific Bioinformatics Conference (APBC 2010)
Decoding HMMs using the k best paths: algorithms and applications
Cheriton School of Computer Science, University of Waterloo, 200 University Avenue W., Waterloo, Ontario, Canada N2L 3G1
BMC Bioinformatics 2010, 11(Suppl 1):S28 doi:10.1186/1471-2105-11-S1-S28Published: 18 January 2010
Traditional algorithms for hidden Markov model decoding seek to maximize either the probability of a state path or the number of positions of a sequence assigned to the correct state. These algorithms provide only a single answer and in practice do not produce good results.
We explore an alternative approach, where we efficiently compute the k paths of highest probability to explain a sequence and then either use those paths to explore alternative explanations for a sequence or to combine them into a single explanation. Our procedure uses an online pruning technique to reduce usage of primary memory.
Out algorithm uses much less memory than naive approach. For membrane proteins, even simple path combination algorithms give good explanations, and if we look at the paths we are combining, we can give a sense of confidence in the explanation as well. For proteins with two topologies, the k best paths can give insight into both correct explanations of a sequence, a feature lacking from traditional algorithms in this domain.