Table 1. Prediction performance for absolute metabolite concentrations in E. coli and C. acetobutylicum under various growth conditions achieved based on: (I) chemical properties [29], (II) mTOW, (III) mTOW when controlling for the variation in measured concentrations that is explained by the chemical properties fit (via Pearson partial correlation [58]), and (IV) via a regression model that combines both methods.

	E. coli				C. acetobutylicum
Pearson correlations (R) for log [predicted concentration] vs. log [measured concentration]	Aerobic Glucose	Aerobic Acetate	Aerobic Glycerol	Anaerobic glucose	Acido-genesis	Solvento-genesis
Chemical properties [29]	0.57	0.41	0.47	0.14	0.16	0.12
	(p<10-9)	(p<10-2)	(p<10-3)	non-significant	non-significant	non-significant
mTOW	0.59	0.61	0.55	0.64	0.46	0.45
	(p<10-5)	(p<10-4)	(p<10-4)	(p<10-4)	(p<10-4)	(p<10-3)
mTOW, controlling for chemical properties	0.48	0.57	0.51	-	-	-
	(p<10-3)	(p<10-2)	(p<10-2)
Integrating mTOW and chemical properties prediction	0.74	0.6	0.58	-	-	-
	(p<10-10)	(p<10-3)	(p<10-3)

Chemical properties were unable to correctly predict metabolite concentrations in E. coli under anaerobic media and in C. acetobutylicum.