search

d-mcgrath / MetaPathPredict

A machine learning-based tool for predicting metabolic modules in incomplete bacterial genomes.
16 stars 1 forks source link

Scrambled results when running on multiple input files #8

Open wichne opened 1 week ago

wichne commented 1 week ago

Hi David, As I spoke about in our meeting yesterday: I ran three cognate MAGs (i.e., the same organism) that were 99, 99, and 91% complete (by checkM) through MPP individually. All three came back with identical MPP profiles. Comparing these to the results from when I did the bulk run (i.e., fed in all the MAG results with a wildcard) (let's call this bulk1) showed that two of the three had very different results (only ~80% of scores matched). So I ran all of the MAGs from just one of the assembly approaches (that included one of the MAGs giving non-matching results) in bulk (bulk2) and then individually (bash loop over the files launching individual MPP runs). The bulk2 results matched exactly the bulk1 results for those MAGs. But the individual results were different. I thought it might be a simple numeric/alphabetic sort error when you do a bulk run, but upon closer inspection that doesn't appear to be the case.

Let me know what additional information you need/want for troubleshooting.

Thanks, Bill

d-mcgrath commented 1 week ago

Hi Bill, Thanks for posting this issue. I'm looking into it now.

Best, David