The use of some multiple-sequence alignments in phylogenetic analysis, part
icularly those that are not very well conserved, requires the elimination o
f poorly aligned positions and divergent regions, since they may not he hom
ologous or may have been saturated by multiple substitutions. A computerize
d method that eliminates such positions and at the same time tries to minim
ize the loss of informative sites is presented here. The method is based on
the selection of blocks of positions that fulfill a simple set of requirem
ents with respect to the number of contiguous conserved positions, lack of
gaps, and high conservation of flanking positions, making the final alignme
nt more suitable for phylogenetic analysis. To illustrate the efficiency of
this method, alignments of 10 mitochondrial proteins from several complete
ly sequenced mitochondrial genomes belonging to diverse eukaryotes were use
d as examples. The percentages of removed positions were higher in the most
divergent alignments. After removing divergent segments, the amino acid co
mposition of the different sequences was more uniform, and pairwise distanc
es became much smaller. Phylogenetic trees show that topologies can be diff
erent after removing conserved blocks, particularly when there are several
poorly resolved nodes. Strong support was found for the grouping of animals
and fungi but not for the position of more basal eukaryotes. The use of a
computerized method such as the one presented here reduces to a certain ext
ent the necessity of manually editing multiple alignments, makes the automa
tion of phylogenetic analysis of large data sets feasible, and facilitates
the reproduction of the final alignment by other researchers.