running on hpc environment

Question

running on hpc environment

Opened this issue 2 years ago · 7 comments

Hi,
I'm running metamorpheus on my cluster with task toml files.
I was wondering if there was a way to add more cores (maybe an extra line in toml file?). Task4 takes several days and the task takes 4 cores when i have 120 to allocate.
Also wondering how to add BayesianFoldChange analysis (flashlfq portion) to my run. Do i add a line in my Task5 toml for that?
TY

Answer 1 · 2023-02-16T17:22:53.000Z

in windows, the "MaxThreadsToUsePerFile" parameter in the toml file adds more cores. This is line 50 in the latest version. Here, I show 7, which is all my laptop will allow.

Answer 2 · 2023-02-16T17:24:33.000Z

Running MM in HPC may not accomodate the use of threads in the same way as for windows. I have almost no experience w/ HPC. Some collaborators worked to adapt MM to HPC. You can check on that here: https://github.com/PSTL-UH/HPCMetaMorpheus I don't think that it is up to date.

Answer 3 · 2023-02-16T17:25:55.000Z

Quantification with FlashLFQ occurs by setting the "DoQuantification" setting in the toml to true (Line 18 in the latest release)

Answer 4 · 2023-02-16T17:27:09.000Z

However, this doesnt do the bayesian thing. You can add a ExperimentalDesign file to the same directory as your raw files and MM will use that in the quant. It allows you to specify conditions/bioreps/fractions/techreps

Answer 5 · 2023-02-16T17:28:39.000Z

Here is an example ExperimentalDesign.tsv file:
ExperimentalDesign.zip

Answer 6 · 2023-02-16T18:13:41.000Z

Thanks for the info. I supplied the Experimental design file and it was read in by metamorpheus. Does that mean I will get the Bayesian fold change file automatically or is there no way to attain that without running flashLFQ separately?

In the case I do have to run FlashLFQ separately after metamorpheus finishes: I set Normalize to "true" in my metamorpheus Task5 toml. Should I still set normalization (with --nor argument) in the FlashLFQ thereafter or leave it out?

Answer 7 · 2023-02-16T19:19:37.000Z

MM will do FlashLFQ and provide quantitative protein values. But it will NOT do the bayesian thing. For that, you have to run FlashLFQ separately. I made an issue to add the bayesian thing to MM. I don't think it will be hard. But it is on a list so it may take a few weeks to complete.
I'm not sure how well normalization works. I would leave it off for now. You can see how it works in FlashLFQ. That has normalization capabilities. I don't think you need normalized data for bayesian anyway. I think that it takes care of what it needs