The relationship between e-value and percent identity

Question

[karn_heera990]

Why are e-values not correlated with percent identity? I did a blastp and found that one alignment has 29.54% identical residues, and an e-value of 2 x 10-129, while another has 29.41% identical residues but an e-value of 3 x 10-21.

[Ram]

Just an aside, your examples don't prove your point. You've given 2 values of 2 variables, and I see negative correlation (not no correlation) between the two. Maybe use more extreme-value examples or provide more examples to establish that there is indeed no correlation?

Also, both matches being at 29% identity, the difference in e-value is probably owing to how complex the region is, how long the match is, etc. It doesn't make sense to just compare percent identity and e values without looking into more factors that influence e values.

[h.mon]

Read the Blast FAQ on e-value, and the link therein for the formulas nitty-gritty.

[Jeremy Leipzig]

show us the alignment - a local alignment could be 100 residues at 29.54% vs 50 residues at 29.41%