No growth with CarveMe v.1.5.0

cdanielmachado / carveme

CarveMe: genome-scale metabolic model reconstruction

Other

145 stars 49 forks source link

No growth with CarveMe v.1.5.0 #122

Closed lumlukwp closed 3 years ago

lumlukwp commented 3 years ago

Hi Daniel,

I built the model like so carve protein.faa —init M9 -u gramneg —cobra -o model.xml. The model I built with v1.2.1 was able to grow on M9 media but the one with v1.5.0 does not. Do you have any suggestions?

Thank you in advance for your help.

cdanielmachado commented 3 years ago

I think this issue is related to this one: https://github.com/cdanielmachado/reframed/issues/2

Can you gap-fill explicitly for M9 and check how many gap-filling reactions are added?

carve protein.faa -g M9 -i M9 -u gramneg —cobra -o model.xml -v

lumlukwp commented 3 years ago

The models were able to predict growth using the gap-filling approach with only few reactions added. However, the growth rate is quite low compare to the previous version. The numbers are found below:

Model	GrowthRate	#Reactions	#Metabolites	#Genes
m9 v1.2.1	0.769926383	2151	1447	1330
m9 v1.5.0	0	2442	1615	1497
m9_gapfill v.1.5.0	0.186136468	2448	1616	1497
LB v1.2.1	2.181619235	2151	1447	1330
LB v1.5.0	0	2442	1615	1497
LB_gapfill v1.5.0	0.370916621	2446	1616	1497

What version should I continue with?

silvtal commented 3 years ago

I have similar results. As I said, the media I was using was M9 (with glucose):

Model	Growth	Reactions	Added
CarveMe 1.4.1	0.796	2296	-
CarveMe 1.5.0	0.0	2468	-
CarveMe 1.5.0 gapfilled	0.120	2471	R_BZDH, R_LO, R_PHPYROX

The 1.5.0 model has 172 more reactions, but when I try a for reactions in cm_150 not in cm_140 it returns 1099 reactions. There are also 927 reactions in the old model that are not in the new one, apparently.

Loading the models with flavor="bigg" returns the same values.

cdanielmachado commented 3 years ago

Thank you both for helping me understand the problem.

Very few gap-filling reactions are added in all cases, this is good. Like I said, without explicitly gap-filling for M9, it is mostly a matter of "luck" if the models will grow on M9 out-of-the-box.
The growth rates are much lower, this doesn't sound so good. Can you send me your fasta files for these models, so I can try to see what is going on?

silvtal commented 3 years ago

Sure. It's from GTDB v95. You can download it here https://gofile.io/d/4teCZ9

cdanielmachado commented 3 years ago

Thanks a lot. I will try to understand what has changed so drastically.

Please feel free to report any other differences that you find between the two versions. Especially differences in phenotype simulation.

cdanielmachado commented 3 years ago

Should be solved in release 1.5.1, otherwise please re-open this issue.