In article <27d2bd1f.0309020613.4eff170b at posting.google.com>,
JamesW <gingaloon at hotmail.com> wrote:
>>I hope someone can help me.
>>A colleague recently asked me whether he should expect different
>evalues from running a blastx using ESTs as the query against a
>protein database or in reverse running a tblastn using the individual
>proteins to seed a search against the EST database.
>>I said yes, as the database size affects the evalue. Which I
>understand. He mentioned that the bit scores were different, which
>after a lot of thought still has me stumped...
>>I presumed that the bit scores would be the same as these are
>calculated in construction of the HSP, which should be the same,
>shouldn't it? Anyways I found examples where it was not. I've pasted
>below the HSPs produced from a blastx and tblastn respectively.
bit scores are normalised with lambda which, as I understand it, depends
on the scoring matrix being used (and presumably also on the gap
insertion/extension penalties being used)
>1.a.)Why is the alignment longer in the latter, when surely this would
>match in the blastx search?
We haven't seen the command line parameters that were used. If they
were the defaults, are you sure that the defaults are the same for
tblastn as they are for blastx, in the particular flavour of BLAST you
>1.b.)From the front of the alignment why is leu-leu identity picked up
>in the first HSP but not the second?
Same here; was the X parameter value the same both times? W and T?
What about the matrix?