fitGAM run time - Githubissues

HectorRDB / condiments

Trajectory inference across multiple conditions with condiments: differential topology, progression, differentiation, and expression

https://hectorrdb.github.io/condiments/

Other

16 stars 1 forks source link

fitGAM run time #11

Closed Sa753 closed 2 years ago

Sa753 commented 2 years ago

Dear Hector,

I am stuck at fitGAM function as it takes >24h to run on a data set of 5000 cells and 36600 genes., 3 conditions and 5 knots. I tried running it on 10 genes and it took 21s then on 1000 genes and it took 1h which would mean that I actually need >24h to run it on all the genes in the count matrix. Is there anything to do for that please? I am sorry if I missed any update that solved this problem. I saw that this issue was raised before on tradeSeq GitHub on
https://github.com/statOmics/tradeSeq/issues/64

Thanks

HectorRDB commented 2 years ago

Hi @Sa753 I understand it can be painful. Unfortunately, there are no easy solutions and your runtimes are reasonable for a dataset of this size. Here are a few possible helps.

Apply filtering. In our experience, the top 10,000 genes are more than enough to capture all relevant biological signal
Use parallelization. See our vignette on this.
Consider running on a cluster computing where jobs can run for more than a couple hours if you have access to those.

Let me know if that helps, Best

HectorRDB commented 2 years ago

closing for lack of activity after one month. Feel free to reopen if issue persists