Ka/Ks Ratio For Within A Gene - Very Confused !!

Question

[User 7433]

Hi,

I am asking another ka/ks related question! I am a bit of a beginner with this stuff so go easy.

I understand the the Ka/ks ratio is used mainly for interspecies comparisons..is a way of assessing what type of selection is acting on a gene.

I know it is also usable for intraspecies comparisons...so different coding sequences of the same gene can be compared to assess what selection is acting on the gene - this is what I would like to do - I have the CDS sequence for >2000 individuals that I have sequenced. Most of them do not have any changes to the coding sequence..as this gene is likely to be essential - no premature stop codons etc have been found.

I plan to do the analysis using DNAsp....do I input the CDS sequence for ALL individuals? Or do I input JUST the 200 or so variable ones and the original CDS sequence..?

Secondly, as you get a ka/ks for each individual sequence...do I average this to assess what's going on overall?

I am pulling my hair out...I have a deadline, I am so confused - any help would be SO appreciated!

Thanks xx

[Alf]

Try to reformulate the question. The fact that you are very confused is understandable because you are here, IMHO.

[a1ultima]

Have a look at the bottom of the best answer on this biostars Q: C: Best Practices/Softwares To Calculate Ka/Ks Ratio

Answer 1

[Vitis]

Within populations or species, there are suggestions that ka/ks may not be a good measure of selective constraints because the selection and ka/ks ratio doesn't follow a monotonic function. See http://www.plosgenetics.org/article/info:doi/10.1371/journal.pgen.1000304 from the Plotkin lab. If I were to use ka/ks here, I'll try codon-based approach implemented in HyPhy to see site and lineage specific dN/dS ratio. Also, I think it might be a better idea to explore population genetics based approach, like Tajima's D, to test the null hypothesis of neutral evolution.

To answer your direct questions, yes, you input all coding sequences, and you do get average values from all pairwise comparisons. But I wonder whether you'll see any significant values because they should be all highly similar.