Dr G. Bunkoczi
gb360 at cam.ac.uk
Fri Nov 19 14:44:48 PST 2010
>i have seen programs where you say e.g. "fubar" and then the alignment
>has a part with ">fubar" then everything between ">fubar" and the next
>">" is defined as the target sequence (IIUC).
OK, this is easy - simply select the sequence with a keyword (which may be
a database record identifier) as target.
>... anyways, how is the target sequence used that a consensus sequence
>couldn't be used instead? i mean, the model looks like it has one
>applied to it.
Many current algorithms only work with pairwise alignments, and therefore
require a target sequence, e.g. delete residues from the model that align
with gaps in the target. Some of these concepts can be generalized to
multiple sequence alignments, and make these decisions based on the local
sequence similarity (calculated from residue substitution scores taking
nearby residues into account), but not all (e.g. if a Phe in the model
aligns with a Gly in the target, one would possibly want the Phe sidechain
to be deleted).
One could possibly get away without a target sequence by selecting the
right algorithms, but I am wondering whether this has any relevance to
current practice. The sequence of the protein is usually known, and indeed
used in the homology search to provide template models.
More information about the phenixbb