How to optimize SNAPP run? Very slow runs even with multithreading

[lncespe...@gmail.com]

Hi all,

I am now trying to run a SNAPP analysis (2,900 SNPs, 10 taxa, 20 individuals) but I am having troubles making it run efficiently. I have tried multithreading, and playing with different values of threads, but it is still very slow. I have a 2'000.000 steps chain running, and after four days not even a quarter is done (still running). I also tried with a shorter run of 200,000 steps but the chain clearly did not achieved stationarity (see screenshot from Tracer attached). 

Is it likely that the run is so slow because the priors are not the best for my dataset? I used uninformative priors for the population size parameters as adviced here: https://evomics.org/learning/population-and-speciation-genomics/2020-population-and-speciation-genomics/species-tree-inference/). Alternatively, I am thinking that the problem might be that I have two few individuals per "species": I only included two thinking that this might make the run faster, but I guess this makes the estimation of some parameters hard? Do you have any suggestions on regards of the number of individuals vs speed of the run?

Any and all advice is super appreciated- I really need to make my run faster to be able to get results in a reasonable time that is under the limitations of my computing cluster.

Thank you!

All the best,

Laura

[higg...@gmail.com]

Hi Laura,

Have you considered using the snapper package (https://github.com/rbouckaert/snapper) instead of the SNAPP package: snapper should work the same as SNAPP but faster and the run time is not sensitive to the number of individuals per species.

Cheers,

Remco

[lncespe...@gmail.com]

Hi Remco,

Thank you so much! The snapper package was not really on my radar, thank you! I think is precisely what I need, and will give it a try with a larger number of individuals.

All the best,

Laura