How is sequence identity affected by non-end gaps?

Question

[n00514450]

From what I read, the sequences

AACTG
AACTGAC

have 100% identity, while the sequences

AACTG
AACTC

have 80% identity. What about

AATCG--C
AATCGAAC

[Brian Bushnell]

I like to use (#matches)/(alignment length), where alignment length includes gaps, because it is symmetric; you get the same identity regardless of which is the query and which is the reference. I count an N as 0.25 matches.

In practice, identity is often not a good metric, because it either gives exaggeratedly low scores to sequences with long indels, or ignores indels completely, neither of which makes much sense.

Answer 1

[Brice Sarver]

There are many ways to calculate percent identity. Some take every character in a DNA string and compare it to another, whereas others explicitly exclude gapped or ambiguous sites. In your third case, the % identity depends on whether or not you include the sites that are gapped in the first sequence.

For a very short yet good synopsis, I recommend this paper by Alex May, "Percent Sequence Identity: The Need to be Explicit."

[n00514450]

Thank you. I guess the best thing to do is just ask my professor how he defines it.