The core of our technology is based on a network-based machine learning algorithm for integrative analysis of untargeted metabolomic data with other large-scale molecular information such as data from genes, proteins, drugs and diseases. Our technology is developed at the Fraenkel lab at MIT Biological Engineering department, and published in Nature Methods.
Among various molecular data, metabolomics provides more functional information. However, existing technologies mostly consider mRNA data and not metabolomics, because of the ambiguity in large-scale metabolomic information. Globally, we can measure metabolite masses inexpensively and fast, but each mass has an ambiguous identity. Only a small fraction of metabolite masses can be clearly characterized via additional time-consuming and costly experiments. Using our proprietary database and machine-learning algorithm, we reduce the need for these additional experiments, predict the identity of each metabolite mass, and integrate these data with other large-scale molecular datasets such as genomics and proteomics.