|
Abstract : |
Sequence weighting methods have been used to reduce redundancy and emphasize diversity in multiple sequence alignment and searching applications. Each of these methods is based on a notion of distance between a sequence and an ancestral or generalized sequence. We describe a different approach, which bases weights on the diversity observed at each position in the alignment, rather than on a sequence distance measure. These position-based weights make minimal assumptions, are simple to compute, and perform well in comprehensive evaluations. Redundancy is a common feature of sequence databanks, where a typical gene or protein family is represented by a highly non-random sample of sequences. For example, an ancient protein family might be represented by a few highly diverged microbial and invertebrate sequences plus many mammalian sequences that form a closely related subgroup. This situation can be detrimental in sequence alignment and searching applications, where it is usually desirable to represent the diversity among related sequences. Since closely related sequences are largely redundant, they provide less information in a multiple sequence alignment than their distant cousins. Sequence weighting methods have been introduced to compensate for over-representation, |
