search

cdanielmachado / carveme

CarveMe: genome-scale metabolic model reconstruction
Other
147 stars 50 forks source link

Model content not reflective of diamond results #66

Closed JaredTBrod closed 4 years ago

JaredTBrod commented 4 years ago

Good day,

I ran CarveMe on a bacterial genome. The .tsv output has about 700 unique genes mapped against the internal GPR database. However, the final model only has ~100 genes. I manually curated some of the missing .tsv hits and they look good, so I'm not sure why they are not in the final model.

Do you have some suggestions as to why all the .tsv content didn't make it into the final model?

I'm running on gurobi as cplex isn't available for me. Not sure if that is relevant.

Thank you,

JTB

cdanielmachado commented 4 years ago

Hi Jared, that is indeed strange. Could you please send me the .tsv output so that I can try to reproduce the results?

JaredTBrod commented 4 years ago

Hi Daniel,

Let's see if this works. I've tried to attach the tsv file here.

Jared

EDIT: Hmm, tsv files won't upload here. Do you have an email I can send it to?

On Thu, Mar 5, 2020 at 12:19 AM Daniel Machado notifications@github.com wrote:

Hi Jared, that is indeed strange. Could you please send me the .tsv output so that I can try to reproduce the results?

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/cdanielmachado/carveme/issues/66?email_source=notifications&email_token=AC5AQPWMQELPVILDOSF2FATRF5OATA5CNFSM4LBZKV7KYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEN4FY7I#issuecomment-595090557, or unsubscribe https://github.com/notifications/unsubscribe-auth/AC5AQPW5KZWW2BQ6PHSL3E3RF5OATANCNFSM4LBZKV7A .

cdanielmachado commented 4 years ago

Yes, my username at gmail.

JaredTBrod commented 4 years ago

I sent it along, please let me know if you did not receive it.

cdanielmachado commented 4 years ago

I did, sorry, just didn't have time to test it yet.

JaredTBrod commented 4 years ago

Ok, no worries, I just wanted to make sure it didn't get lost in the ether. Thank you!

cdanielmachado commented 4 years ago

Unfortunately, it seems to be a problem with Gurobi. Somehow it finds a very different solution from CPLEX for the same MILP problem.

I will try to look into it, but for the moment I must say that we can only officially support CPLEX (where we have done extensive benchmarking of our models).