jbloomlab / phydms

phylogenetic analyses informed by deep mutational scanning data
GNU General Public License v3.0
14 stars 4 forks source link

DMS2_batch_prefs MCMC does not converge on certain ENV sites #16

Closed jiroop closed 5 years ago

jiroop commented 5 years ago

There is a readme file within the zipped folder that describes the issue in more detail and how to reproduce it, but this is the error I am getting when running DMS2_batch_prefs:

2018-11-02 13:35:54,248 - INFO - Getting results for site 206... 2018-11-02 13:35:54,248 - ERROR - Terminating dms2_prefs with ERROR Traceback (most recent call last): File "/home/jroop/.local/bin/dms2_prefs", line 304, in main .format(r, logstring)) RuntimeError: Failed for site 206: Beginning MCMC at Fri Nov 2 11:59:53 2018 After 4 MCMC chains each of 20000 steps, mean R = 1.2541643122083816 and mean Neff = 37.9 MCMC failed to converge. Doing retry 1 with 40000 iterations per chain. After 4 MCMC chains each of 40000 steps, mean R = 1.539564031833223 and mean Neff = 23.35 MCMC failed to converge. Doing retry 2 with 80000 iterations per chain. After 4 MCMC chains each of 80000 steps, mean R = 1.2654302097288201 and mean Neff = 33.2 MCMC failed to converge. Doing retry 3 with 160000 iterations per chain. After 4 MCMC chains each of 160000 steps, mean R = 1.530588807580061 and mean Neff = 57.35 MCMC failed to converge. Doing retry 4 with 320000 iterations per chain. After 4 MCMC chains each of 320000 steps, mean R = 1.2578045547551282 and mean Neff = 151.3 MCMC failed to converge. Doing retry 5 with 640000 iterations per chain. After 4 MCMC chains each of 640000 steps, mean R = 1.5417707240750493 and mean Neff = 199.05 MCMC FAILED to converge after all attempts at Fri Nov 2 13:35:52 2018. 181101_MCMC_debug_short.tar.gz

jbloom commented 5 years ago

@jiroop: This issue should be posted with the dms_tools2 repo, not the phydms repo, since it affects the former.

I have closed issue, can you repost on dms_tools2.