Biomarker Discovery Using New Metabolomics Software for Automated Processing of High Resolution LC-MS Data

摘要

Robust biomarkers of target engagement and efficacy are required in different stages of drug discovery. Liquid chromatography coupled to high resolution mass spectrometry provides sensitivity, accuracy and wide dynamic range required for identification of endogenous metabolites in biological matrices. LCMS is widely-used tool for biomarker identification and validation. Typical high resolution LCMS profiles from biological samples may contain greater than a million mass spectral peaks corresponding to several thousand endogenous metabolites. Reduction of the total number of peaks, component identification and statistical comparison across sample groups remains to be a difficult and time consuming challenge. Blood samples from four groups of rats (male vs. female, fully satiated and food deprived) were analyzed using high resolution accurate mass (HRAM) LCMS. All samples were separated using a 15 minute reversed-phase C18 LC gradient and analyzed in both positive and negative ion modes. Data was acquired using 15K resolution and 5ppm mass measurement accuracy. The entire data set was analyzed using software developed in collaboration between Bristol Meyers Squibb and Thermo Fisher Scientific to determine the metabolic effects of food deprivation on rats. Metabolomic LC-MS data files are extraordinarily complex and appropriate reduction of the number of spectral peaks via identification of related peaks and background removal is essential. A single component such as hippuric acid generates more than 20 related peaks including isotopic clusters, adducts and dimers. Plasma and urine may contain 500-1500 unique quantifiable metabolites. Noise filtering approaches including blank subtraction were used to reduce the number of irrelevant peaks. By grouping related signals such as isotopic peaks and alkali adducts, data processing was greatly simplified by reducing the total number of components by 10-fold. The software processes 48 samples in under 60minutes. Principle Component Analysis showed substantial differences in endogenous metabolites levels between the animal groups. Annotation of components was accomplished via searching the ChemSpider database. Tentative assignments made using accurate mass need further verification by comparison with the retention time of authentic standards.