January 24, 2018
Potential to identify geographical origins and seasonal deliciousness of food
RIKEN CSRS has developed a deep-learning (DL) algorithm optimized for metabolomics research. The algorithm was then used to analyze NMR data for fish, demonstrating that it is possible to identify geographic origin with a high level of accuracy and establishing a method to determine important metabolites that contribute to such identification.
Artificial intelligence (AI) is an innovative technology that will contribute to the transformation of social structures. AI and DL also have potential in biology and chemistry, but they have yet to be applied in large measure.
CSRS researchers developed a DL algorithm based on a deep neural network (a principal calculation algorithm for DL) that enables to identify important metabolites contributed to classification/regression models in metabolomics research. The researchers applied the DL algorithm to a data set of more than 1,000 metabolic profiles from fish muscle extracts obtained by NMR measurements (which are suitable to big data acquisitions due to highly reproducible and compatible between laboratories) with comparisons of conventional discriminant analysis and several machine learning approaches. The results demonstrated that the DL algorithm was able to determine the geographical origin of the samples with the greatest accuracy and, furthermore, could identify the important factors (metabolites in the case of this research) that contributed to geographical determination.Looking to the future, the DL algorithm in combination with basic and cost-effective equipment should be applied as quality control tool for agricultural and fishery products, contributing to the development of valuable food products related to seasonality and geographic origin based on determination of important factors as metabolic markers.
Analytical Chemistry doi:10.1021/acs.analchem.7b03795
Y. Date, J. Kikuchi,
"Application of a deep neural network to metabolomics studies and its performance in determining important variables".
Jun Kikuchi; Team Leader
Yasuhiro Date; Research Scientist
Environmental Metabolic Aanalysis Research Team