aBSREL run time

[lagphase]

Hello Sergei,

I am running aSBREL on a 21 Mb alignment, hyphy version 2.5.36. The analysis has been running for over 1 day and it's still at

"Fitting the baseline model with a single dN/dS class per branch, and no site-to-site variation."

Do you know approx. the run time for this alignment size? Do you think a newer hyphy version would speed up the process? Thanks.

[spond]

Dear @lagphase,

1. I would suggest updating to the latest HyPhy version. We did a lot of work in the recent 2+ years to improve speed.
2. How many sequences are in the alignment?

Best, Sergei

[lagphase]

Hello Sergei,

Thanks for your prompt response. The analysis has stopped with errors.

I have 12 sequences in my alignment (attached below). Could you take a look and let me know what went wrong? Thank you. pal2nal-12-seqs.fas.gz

> Error:Internal error 
>  
>   ERROR: [_LikelihoodFunction::LocateTheBump (index 65) current value   -25775513.04835993 (parameter =   0.9992685062681751), best value   -25775513.04835992 (parameter =   0.9992685062681751)); delta = 3.725290298461914e-09, tolerance =                
> 
> Parameter name NGYEXnlA.tree_0.N4.delta
> Internal error 
>  
>   ERROR: [_LikelihoodFunction::LocateTheBump (index 65) current value   -25775513.04835993 (parameter =   0.9992685062681751), best value   -25775513.04835992 (parameter =   0.9992685062681751)); delta = 3.725290298461914e-09, tolerance =                
> 
> Parameter name NGYEXnlA.tree_0.N4.delta

[lagphase]

Hello Sergei,

Based on previous issues, it seems like if I add ENV="TOLERATE_NUMERICAL_ERRORS=1;" in the command line, the problem would be solved??

I initially used the interactive prompts hyphy -i. I updated hyphy to the newest version and restarted the analysis using command line with added the line above. Hope this will work this time.

[lagphase]

Hello Sergei,

The analysis succeeded in ~ 8 hours., which was fantastic!! I have a question regarding aBSREL results which indicated that the branches I labeled as Foreground are under positive selection. But is this result relative, e.g. does this mean other unlabeled branches (background) are not under selective pressure? Or the method completely ignore other background branches?

[spond]

Dear @lagphase,

aBSREL does not do any testing on background (not selected branches). They could be under any kind of selection, including positive selection.

Also, in v 2.5.62, I added an option to aBSREL to support bag of little bootstrap upsampling approach (see https://pubmed.ncbi.nlm.nih.gov/34734192/ for more details). You can specify, --blb 0.7 and run aBSREL 10 or so times, each time with a different random sample of your data. Because of what blb does, each analyses should be much faster than the full original analysis. You can then average, or take majorities, over replicate results.

Best, Sergei