[Bioperl-l] Re: Comparative genomics - regions of synteny and whole genome duplication

Jim Kent jim_kent@pacbell.net
Thu, 04 Oct 2001 10:48:45 -0700

>  > You could get some clustering from large scale duplication
>  > events before the species split.  Consider the 4 vertebrate
>  > hox clusters.
> Absolutely, but within species with 4 clusters, wouldn't you still expect the 
> orthologous clusters to be more similar to each other than the paralogous 
> clusters?

Yep.   The key here though isn't a particular scoring threshold though
so much as looking for the best match (presuming you have the
sequence of the whole genome).    Since the rate of sequence divergence
varies pretty wildly across loci,  I'm not sure what to do to discriminate
paralog from ortholog if you lack the whole genome.   Even when you do have
the whole genome there will be cases where the ortholog is lost but the 
paralogs live on.   

Still if you're satisfied with being right 95% of the time or so, a combination
of good blast scores and genes in the same order in both species
would do the trick I'd think.