To generate benchmark data for assessing and correcting inter-batch biases in untargeted metabolomics, we created a set of 136 samples with diverse combinations of matrix (water or serum), spike-ins, and dilutions. Samples were aliquoted, frozen, and analyzed repeatedly in seven independent batches over a period of two months. Each batch encompassed >700 sample injections including technical triplicates and intercalated blanks. Each sample ...