The human body contains hundreds of thousands of naturally occurring chemicals, and is exposed to hundreds of thousands more during daily life. These chemicals, called metabolites, are involved in key biological processes in humans and other large organisms. Metabolomics is the scientific discipline that allows scientists to measure these small molecules at a large scale. Metabolomics can reveal molecules involved in disease, produce diagnostic and prognostic tests and predict how patients will respond to specific prescription drugs. Much has been accomplished with current computational tools, which can only identify a small fraction of the metabolites in a sample. Better computational tools that could identify the remaining metabolites would dramatically accelerate the pace of metabolomics research.
Drs. Leonard Foster of UBC and David Wishart of the University of Alberta are developing computational tools based on an artificial intelligence technique known as “deep learning” to handle the huge amount of data generated by metabolomics experiments. The first tool, DeepMet, will increase the number of molecules that can be identified in metabolomics experiments. The second, MetUnknown, will help assign chemical structures to molecules that are as of yet unknown. Together, these tools will help shine a light on the majority of the metabolome that is overlooked by current tools.
DeepMet and MetUnknown will be available in three different formats so that scientists in many different areas can use the tools without any specialized training. Ultimately, these tools could help researchers identify therapeutic targets for complex diseases such as cancer and develop new tests to help physicians personalize medical treatments.