A heuristic algorithm for multiple alignment was developed to more effectively study alternative splicing patterns of eukaryotic genes. Segment to segment alignment is guided by a guide tree. The time for constructing the guide tree is reduced remarkably by adopting a similarity estimation model which estimates the similarities between sequences from the total number of hits. The segment to segment alignment eliminates the problem introduced by the "gap" which exits in traditional alignments so that alternative splicing patterns are revealed more precisely. The new method for constructing the guide tree and the basic local alignment search tool increase the algorithm speed relative to other multiple alignment algorithms. This algorithm is a new way to study alternative splicing.