I am trying to calculate Ka/Ks ratios for a bunch of exons between different species. What is the minimum sequence length needed for the value of this ratio to be significant and pass the likelihood ratio test (LRT)?

Please encourage me and my buddy, we would like to win the car!

In theory, you might get a Ka/Ks significantly > 1 when comparing shorter sequences, if indeed the true Ka/Ks is high (signature of positive Darwinian selection). For example, try running your software (PAML?) on two imaginary sequences that are 15nt long and have only one synonymous change between them, and 4 nonsynonymous changes.

Ryan

FYI:
two 45 nt sequences:

TGGCAATCCGGTGAATGGCAATCCGGTGAATGGCAATCCGGTGAA

GGGAAAACCAGTGAAGGGAAAACCAGTGAAGGGAAAACCAGTGAA

When I do the LRT I get a Ka/Ks significantly > 1 (I put 3 synonymous changes between the two, the rest are non-synonymous). Though, in practice, a Ka >>Ks is generally not this extreme!

Ryan