Solid organic waste represents an abundant, cheap, and available source of biodegradable substrates not yet exploited to produce biohydrogen by dark fermentation. The impact of the composition of solid organic waste on microbial metabolic pathways and subsequently on biohydrogen production, has not been clearly elucidated. The aim of this study is to determine the compositional features of different substrates that influence bioH2 production. For this, we measured Biological hydrogen potentials (BHP) on 26 different substrates and performed a multivariate statistical analysis of the experimental data using a partial least square regression. The results showed that the BHP values correlated well with the initial carbohydrate content measured after mild hydrolysis. A predictive model explaining more than 89% of the experimental variability was then built to predict the maximal biohydrogen yield with a high accuracy and for a large spectrum of organic waste. An explicative model showed that only carbohydrates, butyrate and lactate concentrations were significant variables explaining more than 98% of biohydrogen yield variability. Interestingly, an interaction term between carbohydrates and lactate concentrations was required to explain microbial pathways producing hydrogen.