Hi everyone,
I would like to increase the speed of STRUCTURE software by splitting the work on two computers or two instances running simultaneously. I want to run for 10 clusters.
Therefore, I would like to work with the same data but instead of searching for K 1 to K 10 I split the job on two instances of the STRUCTURE software. One running for K 1 to 5 and the other one running for K 6 to 10. Again: the input data and all other settings will be exactly the same for both instances of the STRUCTURE software.
I've tried it with a small dataset. The clustering of the merged dataset was the same as the original data.
I know that I can speed up STRUCTURE by reducing the number of Burnin or MSMC reps. But that’s another issue.
I would be happy to get feedback from the community if my approach is appropriate or giving me wrong results.
Thank you.
Said
Is using fastSTRUCTURE an option for you?
Article fastSTRUCTURE: Variational inference of population structure...
Thank you Jose for your reply. Actually, not really. I am not that happy about the result I get by using fastSTRUCTURE. Besides, it is not running on Windows.
I found there are some tools like ParallelStructure. But it was not easy to use.
So I splitted the data and merged the output file. That works well for me.
Said
As far as I understood, the approach is as valid as testing 1
- Badges
- Science topic