Help with selecting alpha in codeML

[Saar Koene]

Hi! i am quite a novice user of PAML. I am doing an ancestral sequence reconstruction. I am confused about the parameter alpha. I know it is related to the gamma distribution which refers to the substitution rates of the amino acids. But my understanding does not go much deeper. I ran two reconstructions: (1) with fix alpha enabled and alpha=0 and (2) with fix alpha disabled and alpha=0.5. 

The differences in the resulting ancestral sequences are subtle but present. How do I decide which model is better/if I need a different alpha value? 

Any recommended literature is also appreciated!

Thanks in advance!

[Sishuo Wang]

Hi Saar,

Thx for the question! Typically you needn't choose alpha. If not fixing alpha, then a best fit would be selected by paml, no matter whether you specify the value of alpha. Having it automatically determined is suggested, unless you'd like to check the impact of other values (thus alt hypotheses) on the result. see http://treethinkers.org/tutorials/the-gamma-distribution/ for some more details on the background.

best,

sishuo

[Saar Koene]

Hi Sishuo,

Thank you so much for your fast reply! I think I get it now. I found that the reconstruction where I did not fix alpha indeed showed a different alpha than 0.5. So now I get that this is because PAML selected the best fit. The tutorial you sent also is usefull, thank you!

Kind regards,

Saar

Op dinsdag 1 april 2025 om 04:56:03 UTC+2 schreef Sishuo Wang: