Abstract
In MEXT Program for Scientific Research on Innovative Areas “Systems Cancer” and “Systems Cancer in Neo-Dimension” (2010-2019), we developed a large-scale genome data analysis pipeline called Genomon in collaboration with Professor Seiji Ogawa (Kyoto University). Our efforts successfully produced innovative results on cancer genomics. This system is implemented on the supercomputers SHIROKANE and FUGAKU. One of the contributions unraveled the overall picture of genetic abnormalities in malignant brain tumors (Mutational landscape and clonal architecture in grade II and III gliomas. Nat Genet 2015) that exploited Genomon on SHIROKANE. However, with the spread of new measurement technology and new computing environments, no one thinks that the future can be figured on this simple extension. On the other hand, for cancer genomic medicine, Institute of Medical Science University of Tokyo made a research team analyzing whole genome sequences. The challenge we faced was to transform thousands to millions of genomic aberrations per case into precision medicine. It is what we now call “digital transformation.” IBM’s Watson for Genomics was introduced for our research purpose. In the process, we identified the effectiveness of AI, the indispensability of specialist intervention, and bottlenecks. We recognized that natural language processing technology such as BERT and Google Knowledge Graph AI technology will open up the future. Automatic document creation is also a realistic issue. Cancer research is getting more difficult and larger in scale. For example, analysis of genomic data from 60, 954 cases revealed a new underlying mechanism in which multiple mutations within the same oncogene synergistically work (Nature 2021). AI with an accuracy of X% does not seem to be the goal. What is needed is not a black box, but explainable AI that explains “why” in a human-understandable way. We are currently conducting research with Fujitsu Laboratories for this direction.