A Groundbreaking Method for Analyzing Omics Data Created by an Unconventional Biomedical Scientist

An unconventional background, comprising studies in elementary particle physics, information science, and medicine.
Leveraging his diverse knowledge and experience, he has developed an innovative method for omics analysis.

Research into the comprehensive analysis of molecules in living organisms is called omics. Omics research spans a variety of targets: genomics for genetic information (genomes), proteomics for proteins (proteomes), and metabolomics for metabolites (metabolomes). All these approaches have the common aim of elucidating the mechanisms of life phenomena and diseases.

Professor Tatsuhiko Tsunoda has developed a powerful tool for omics analysis: DeepInsight, an artificial intelligence technique for advanced image processing, that is capable of classifying omics data with unprecedented accuracy.

One of Professor Tsunoda’s research interests is the study of oncogenes, the genes that cause cancer.

“If we can classify the type of cancer based on omics data, for example those of oncogenes, we can select the optimal treatment for the patient. However, there are numerous omics patterns associated with cancers, and the patient population for each is very small. Hence, conventional statistical analysis has its limitations.”

Professor Tsunoda therefore turned his attention to deep learning techniques for handling nonlinear data such as images.

Deep learning excels at image processing, but cancer omics data themselves are not composed of image data. Professor Tsunoda thought that if omics data could be treated like image data, deep learning could be applied to omics analysis.

“I liked math and knew there were ways to compress higher dimensions into lower dimensions. Omics data are with high-dimensional variables, while images are typically two-dimensional. I was inspired by the idea that if we could represent omics data in two dimensions, they could be handled as image data.”

This conversion of omics data to image form is the key point of his research. The two-dimensional omics data are then applied to deep learning – one of the artificial intelligence techniques. This enables highly accurate data classification, far superior to that with typical machine learning techniques.

Professor Tsunoda says that the DeepInsight method is universally applicable.

“In principle, it is possible to apply deep learning to any data, including omics data, provided that it is first converted to two dimensions. In fact, we are able to accurately classify not only omics data for cancer, but also text data and voice data. In the future, we aim to further improve our deep learning capabilities to be able to handle more complex data combining multiple omics approaches.”

Although Professor Tsunoda now researches life and medical sciences, his background is very diverse. He studied elementary particle physics up to the end of his master’s program, and then entered the world of information science for his doctorate. After doctoral studies in engineering, he also earned a doctorate in medicine and tackled the analysis of big data in medicine. It is this exceptional background that allowed the professor to develop the DeepInsight method.

“This background gave me a flexible outlook, he said. Even though the work was in different fields, the underlying science is common to all.”

It’s a way of life as a researcher, whether you deepen one interest or tackle several interests. Some things cannot be created without having a diverse perspective.

Interview and text: Jiro Urushihara
Editing: Masatsugu Kayahara
Photography: Junichi Kaizuka

Originally published in The School of Science Brochure 2020

TSUNODA Tatsuhiko

Professor, Laboratory for Medical Science Mathematics, Department of Biological Sciences, School of Science / Department of Computational Biology and Medical Sciences, Graduate School of Frontier Sciences, The University of Tokyo.

In 1995, he completed the doctoral program in the Graduate School of Engineering at the University of Tokyo. In the same year, he became an assistant professor at the Graduate School of Engineering, Kyoto University, then a research associate at the Institute of Medical Science, the University of Tokyo in 1997, an assistant professor at the same institute in 1998, and a team leader at the RIKEN SNP Research Center in 2000. He has been in his current position since 2019.