Closed ChadFibke closed 2 years ago
Hi, thanks for asking about this case.
It looks to me like none of the GMM, DBSCAN or refine models would fit well to this data, based on that distance distribution plot. There aren't distinct components, and specifically not one near the origin separated from others.
First thing to check - are these core/accessory distances what you expect? What species are you looking at. Is it perhaps within a strain, or a less diverse virus?
I think there are three things you could try here:
--indiv-refine
and setting an appropriate --manual-start
(then using the 'core only' fit).Hi @johnlees,
I'm working with SARS-CoV-2. The input are consensus sequences (superimposing called variants onto the Wuhan reference). I did not expect to see a lot of variation, and was surprised there was variation in the accessory.
Thank you for all the options! I successfully ran the lineage model, and it seemed to have satisfied our purposes (to control for population structure in an elastic net model). I will also look into PopPIPE!
Best, Chad
Great! Glad it worked
Hi @johnlees, Thanks for making this interesting program! I'm currently running into some issues stated below:
Versions
I ran the following command to generate the database:
Describe the bug
This constructed all the expected files and the following distance distribution plot:
I then fit a model to the database using the following command:
The following output was received:
I checked the output location and all of the expected files were writen to file and seem complete (_clusters.csv, _unword_clusters.csv, _graph.gt). In addition the following HDBSCAN cluster plot was provided:
There appears to be poor clustering in the initial model, so I'm currently running a
poppunk --fit-model refine
command.However, I was wondering if the recursion issue is interfering with the integrity of the clustering and whether it is safe to continue with refining the model?
Best, Chad