We have analyzed a total of 12 different global and local multiple pro
tein-sequence alignment methods. The purpose of this study is to evalu
ate each method's ability to correctly identify the ordered series of
motifs found among all members of a given protein family. Four phyloge
netically distributed sets of sequences from the hemoglobin, kinase, a
spartic acid protease, and ribonuclease H protein families were used t
o test the methods. The performance of all 12 methods was affected by
(1) the number of sequences in the test sets, (2) the degree of simila
rity among the sequences, and (3) the number of indels required to pro
duce a multiple alignment. Global methods generally performed better t
han local methods in the detection of motif patterns.