dNdS, parameters: codeml sites model

[Amber Nashoba]

Despite ongoing efforts, I am confused regarding output file dNdS values and parameter values reported for codeml site models. I am using 0,1,2,7,8,8a.

I would be really grateful if someone could explain meaning/interpretation/inference of each model's parameter set (multiple p, multiple w, q, etc) and dNdS. This includes "w" values in relation to dNdS values.

 

If there is already a resource please share. I have been through most of the paml literature.

[Sishuo Wang]

Hi Amber,

I guess u were wanting to know that about the site models.

Note as follows for the notations i use w0 to represent omega0 for simplicity and so forth.

m1a: nearly neutral; where a proportion of p0 of the sites have the same dn/ds=w0<1, while the rest (1-p0) proportion have dn/ds = 1. so 2 params to estimate, p0, and w0

m2a: where p0 of the sites have the same dn/ds=w0<1, p1 of the sites have dn/ds=w1=1, while the rest (1-p0-p1) dn/ds > 1, so 4 params to estimate p0, p1, w0, w2

m7: where all of the sites have dn/ds=w0<=1 but they do not necessarily have the same value but instead follow a beta distr parameterized by 2 parameters a and b, thus Beta(a,b), in total 2 params a, b. btw typically to numerically calculate the integral of a beta pdf in paml, it's divided into 8 classes of equal prob in a way similar to perhaps more famous discrete gamma but where 4 categories are often used originally developed by (yang 1994)

m8:  where p0 of the sites have dn/ds=w0<=1 which follows Beta(a,b), and 1-p0 of the site with dn/ds = ws > 1, so in total 4 params (p0, a, b, ws)

m8a: similar to m8 but but w=1 instead of ws>1 (not very sure)

u can refer to Chapter 11 from Yang's MESA (yang 2014), pamldoc, or a recent paper by sandra, kapli & yang https://academic.oup.com/mbe/article/40/4/msad041/7140562, but next time u might want to make ur question more specific or more clearly defined instead of a very general question which can be found in many resources included the above 3.

best,

sishuo

[Sishuo Wang]

sorry, the following is wrong:

"the integral of a beta pdf in pamls"

it should be corrected to: the integral of the phylogenetic likelihood assuming w0 ~ Beta(a,b)

best,

sishuo