Unbiased metabolomic surveys are used for physiological, clinical and genomic studies to infer genotype-phenotype relationships. Long term reusability of metabolomic data needs both correct metabolite annotations and consistent biological classifications. We have developed a system that combines mass spectrometric and biological metadata to achieve this goal. First, an XMLbased LIMS system enables entering biological metadata for steering laboratory workflows by generating ‘classes’ that reflect experimental designs. After data acquisition, a relational database system (BinBase) is employed for automated metabolite annotation. It consists of a manifold filtering algorithm for matching and generating database objects by utilizing mass spectral metadata such as ‘retention index’, ‘purity’, ‘signal/noise’, and the biological information class. Once annotations and quantitations are complete for a specific larger experiment, this information is fed back into the LIMS system...