The identification of active ingredients in crude plant extracts offers great advantages. In this study, nuclear magnetic resonance and chemometrics were used for the screening of in vitro anti-TNFα activity in different berry types. Solid phase extraction was applied and the resulting water, methanol-water (1:1), and methanol fractions were tested for the activity. The methanol-water fraction contained most of the phenolics and showed significantly higher activity than the other two fractions. In the second phase of this study, grapes from 'Trincadeira', 'Touriga Nacional', and 'Aragonês', at four developmental stages were metabolically classified and tested for the TNFα inhibition. The initial stages of grape development, green and veraison, were found more active against TNFα production as compared to the later ripe and harvest stages. Among the cultivars, 'Touriga Nacional' was found to be the most potent inhibitor. Different multivariate data analyses algorithms based on projections to latent structures were applied to correlate the NMR and TNFα inhibition data. The variable importance in the projections plot showed that phenolics like quercetin, myricetin, (+)-catechin, (-)-epicatechin, caftarate, and coutarate, were positively correlated with high activity. This work demonstrates the great potential of NMR spectroscopy in combination with chemometrics for the screening of large set of crude extracts, to study the effects of different variables on the activity, and identifying active compounds in complex mixtures like plant extracts. ELECTRONIC SUPPLEMENTARY MATERIAL: The online version of this article (doi:10.1007/s11306-012-0406-8) contains supplementary material, which is available to authorized users.