-
Notifications
You must be signed in to change notification settings - Fork 37
Amino-acid models not implemented correctly? #130
Comments
Yeah, so the WAG model basically specifies a symmetric matrix for use in the GTR. Just like the GTR, you need to add amino-acid frequencies to get a rate matrix: Q(i,j) = symmetric(i,j)*pi[j]. I believe the WAG matrix comes with a set of frequencies -- so you could use those. But using the symmetric matrix as the Q matrix is only right if the amino acids have equal frequencies, which they don't. So, it looks like the WAG model is not implemented correctly. It should be pretty easy to fix, though. |
@mlandis Michael, am I right that you coded up the WAG and LG functions? Would you be interested in working on this during the hackathon? |
Looking at |
Hey @bredelings, sorry to be slow to reply. I was out of town this past week. Thanks for identifying this mistake -- a good catch! Empirical rate matrices were not correctly multiplying state columns by the stationary frequencies. This appears to have been an old bug that was propagated into newer empirical matrices. Commit 5e5c494 should fix this issue for the following classes: RateMatrix_Blosum62 For cleanliness and consistency, we should assign values to I'll close this issue. We can open a new one if needed. Here are the new rate matrix values:
|
Hi,
It looks like the fnWAG and fnLG models returns symmetric matrices. However, the Q matrix for the WAG isn't symmetric. Am I missing some kind of function that combines the WAG S-matrix with frequencies to get a Q matrix?
-BenRI
The text was updated successfully, but these errors were encountered: