1
$\begingroup$

I want to look at violation of the strict molecular clock and compare dn/ds ratio to test differences in evolutionary rates in a set of genes. Some of these genes are really short (30AA), is it likely to prevent the detection of rate changes ?

Thanks

Improve this question
$\endgroup$
5
  • $\begingroup$ If your genes are really short you may want to find a mathematical hack to get a confidence interval (several published), otherwise the jitteriness will be too much. $\endgroup$
    – Matteo Ferla
    Commented Dec 16, 2019 at 11:12
  • $\begingroup$ I can't find anything about size correction/confidence interval in this type of analysis. As dn/ds is a ratio, why would it be biaised with shorter genes? I feel this can be an issue but I can't find anything explianing why and how to deal with that. $\endgroup$
    – LauraR
    Commented Dec 16, 2019 at 13:22
  • $\begingroup$ Would concatenating several genes make the trick? For example, based on the evolutionary model that fit them the best (concatenate all genes under GTR+I model for example) or all genes related to the same function? $\endgroup$
    – LauraR
    Commented Dec 16, 2019 at 13:34
  • 1
    $\begingroup$ It is not a bias (no size correction needed), it is just that a ratio of two integers gets 'jittery' with low values (say 1/1 = 1, 1/2 = 0.5!). Concatenating different genes would assume the selection pressures are the same, so that would be not advisable. If I am completely misunderstanding and you were going to average (or similarly process) these anyway, averaging them will get rid of the jitters. In terms of errors the simplest and most rudimentary calculation is a margin of error, but that is pointless if the values are to be combined. $\endgroup$
    – Matteo Ferla
    Commented Dec 16, 2019 at 13:55
  • 1
    $\begingroup$ Ok thanks. I'm not willing to combined the values if it's avoidable. $\endgroup$
    – LauraR
    Commented Dec 16, 2019 at 14:37

0

Reset to default

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.