Korean team wins NCI-CPTAC DREAM Proteogenomics Challenge for the first time
Professor Jaewoo Kang’s research team ranks highest in the NCI-CPTAC DREAM Proteogenomics Challenge, a global precision medicine competition to predict the characters of the proteome of tumors.
The results of competition will be published in Nature Methods.
A research team led by Professor Jaewoo Kang of the Department of Computer Science and Engineering, College of Informatics, wins first place in the global precision medicine competition.
DREAM Challenges, which started in 2007, are a collaborative intelligence research community where researchers from all over the world cooperate and compete in order to solve the toughest problems in biomedicine. The findings from the DREAM Challenges were published in prestigious academic journals such as Nature, Cell and Science. Leading research institutions across the world such as National Institutes of Health (NIH), the Dana-Farber/Harvard Cancer Center, UK’s Sanger Institute, and IBM Research have sponsored and organized these Challenges.
This is the first time a Korean team took first place in the DREAM Challenge. Professor Kang’s team also took second place in the AstraZeneca-Sanger Drug Combination Prediction DREAM Challenge held in 2016.
In this DREAM Challenge, teams had to solve three questions on predicting activation levels of proteins in ovarian and breast cancer patients, using the data sets generated by National Cancer Institute’s Clinical Proteomic Tumor Analysis Consortium (NCI-CPTAC).
Tumors occur when a mutation in a gene effects the activation of proteins. These affected proteins are factors of tumor cell growth and metastasis, and may even induce anti-cancer drug resistance. Therefore, more effective cancer treatment plans personalized to each patient’s characteristic can be devised by analyzing and measuring proteins. Technologies measuring the activation level of proteins are developing rapidly. However, discovering links between vast genomic data (DNA and mRNA) and measured proteome data, and then deducing relevant information that may increase the effectiveness of cancer therapeutics remains challenging.
Prof. Kang’s team developed an algorism, based on machine learning technology, which utilizes the genomic information of cancer patients in order to predict the activation of proteomes. At the same time, the algorism predicted the activation of proteomes that were not measured in the data set based on the volume of other proteins found in the cancer patients. Artificial intelligence with significant accuracy in predicting protein volumes from not only given data sets, but also drawing from publically known bioscience knowledge was successfully implemented in this process.
Professor Kang expressed that “An artificial intelligence applying machine learning medical big data found the clue to understanding the life activity of cancer cells. I hope this research can open up new roads to proteome analysis, since existing cancer research mostly focused on genome analysis.”
Competing with around 60 world-class research teams, Kang’s team took first place in Sub-challenge 1 and second place in Sub-challenge 2, despite having only computer science majors as team members (Sunkyu Kim, Heewon Lee, Gunwoo Kim, and Hwisang Jeon, Minji Jeon, Yonghwa Choi, and Daehan Kim). The team from UCLA team ranked 3rd (Sub-challenge 1) and the Stanford University team ranked 13th (Sub-challenge 2).
The final results of the NCI-CPTAC Dream Proteogenomics Challenge will be published in Nature Method, a renowned scientific journal in this field.