Reader Comments

Post a new comment on this article

Nice paper, but...

Posted by tomk on 11 Aug 2011 at 09:44 GMT

This is a nice paper that I certainly will use in my introductory bioinformatics course. But the description on p. 4 about how the various BLOSUM matrices were constructed is not completely correct. The authors state that "BLOSUM62 was created from an alignment of sequences with at least 62% sequence identity, while BLOSUM80 used an alignment with 80%." Actually, BLOSUM62 was created from alignments of sequences in which sequences with more than 62 % sequence identity were clustered and where each cluster was weighed as a single sequence in counting pairs. For BLOSUM80, sequences with more than 80 % sequence identity were clustered. In this way, the contribution of closely related segments to the frequency table was reduced. So the approach of Henikoff and Henikoff was, to put it bluntly and maybe a bit imprecisely, almost the opposite of what is described in the present paper.

No competing interests declared.