Compute Blastx E-Value

Question

[snayfach]

Does anyone know how blastx e-values are computed? Is it the same as blastp, except the query sequence is divided by 3? I've tried using E = (m/3)n2^(-S), but I get slightly different results than from the blastx output. I need to compute blastx e-value because my blast database is split up into several parts.

Thanks in advance

Answer 1

[Bill Pearson]

I believe the blastx e-value uses 2*m, not m/3, because all 6-frames are included.

Answer 2

[snayfach]

I don't understand why, but this formula actually got me pretty close (well under an order of magnitude difference):

E = (m/6)(n)(2^-S)

-m is the sequence length in nucleotides
-n is the database size in residues
-S is the bitscore

It should be noted that blast performs a "finite size correction" in which a value is subtracted from the query and database sequence length. It's unclear what this value actually is.

http://blast.ncbi.nlm.nih.gov/Blast.cgi?CMD=Web&PAGE_TYPE=BlastNews