Within non-communicable diseases, chronic inflammatory conditions represent one of the biggest challenges for modern medicine. Traditional Chinese Medicine (TCM) has been practiced over centuries and has accumulated tremendous empirical knowledge on the treatment of such diseases. Huangqi Jianzhong Tang (HQJZT) is a famous TCM herbal formula composed of Radix Astragali, Ramulus Cinnamomi, Radix et Rhizoma Glycyrrhizae Praeparata cum Melle, Radix Paeoniae Alba, Rhizoma Zingiberis Recens, Fructus Jujubae and Saccharum Granorum (maltose), which has been used for the treatment of various chronic inflammatory gastrointestinal diseases. However, there is insufficient knowledge about its active constituents and the mechanisms responsible for its effects. The present study aimed at identifying constituents contributing to the bioactivity of HQJZT by combining in vitro cytokine production assays and LC-MS metabolomics techniques. From the HQJZT decoction as well as from its single herbal components, extracts of different polarities were prepared. Phytochemical composition of the extracts was analyzed by means of UPLC-QTOF-MS/MS. The inhibitory effects of the extracts on TNF-α, IL-1β and IFN-γ production were studied in U937 cells. Phytochemical and pharmacological bioactivity data were correlated by orthogonal projection to latent structures discriminant analysis (OPLS-DA) in order to identify those HQJZT constituents which may be relevant for the observed pharmacological activities. The investigations resulted in the identification of 16 HQJZT constituents, which are likely to contribute to the activities observed in U937 cells. Seven of them, namely calycosin, formononetin, astragaloside I, liquiritigenin, 18β-glycyrrhetinic acid, paeoniflorin and albiflorin were unambiguously identified. The predicted results were verified by testing these compounds in the same pharmacological assays as for the extracts. In conclusion, the anti-inflammatory activity of HQJZT could be substantiated by in vitro pharmacological screening, and the predicted activities of the OPLS-DA hits could be partially verified. Moreover, the benefits and limitations of MVDA for prediction pharmacologically active compounds contributing to the activity of a TCM mixture could be detected.