Approved research

Prediction of multi-biomarkers for various diseases using deep learning platform

Lay summary

The proposed project is to identify multi-genetic biomarkers for various diseases using deep learning platform. We have aims to determine unique pattern of multi-biomarkers on different cancer types (or other disease forms) in UK biobank data and to evaluate the biomarkers on Korean population, to establish the prediction model of drug responses based on subtypes of disease and on patients and apply to patient stratification and finally to validate and improve our deep learning platform by retraining it for an independent clinical data. Recent advent on personalized medicine requires a precise analysis on a large data set to generalize the prediction models but to specify precise targets. The development of new therapeutics is a challenging process with high costs and failure rates. Particularly clinical trial is the most time and cost-consuming step and also critical step to achieve the ultimate personalized medicine. Predictable biomarkers associated with disease will enhance the process of patient stratification by excluding high risk patients and predicting good responders and increase the efficiency of clinical trial. Accumulating wealth of technology on collecting good quality of genetic and electrical health data and on applying deep learning to biomedicine, gradually makes this possible finally to the precision medicine. We have our own developed platforms to analyze the variants from calling variants and to identifying biomarkers. Genetic biomarkers that sensitize individuals from disease treating drug will be identified and prioritized for further validation to understand the biological mechanism. Understanding biomarkers will improve our ability to develop effective trials during drug development for human diseases. In our scope to achieve this goal, artificial intelligence platform we developed will be applied to identify multi-biomarker for particular clinical associate that model specific aspects of human disease by using the genetic data and corresponding health data from UK Biobank accompanied with other data, TCGA, Korean's genetic/ clinical data. The estimated duration of the proposed project will be 3 years. Our efforts will provide sets of biomarkers customized based on drugs lined up in developmental pipeline to improve the patient subgrouping. As another expected value with UK biobank, we can enhance the accuracy of our platform to predict disease associated biomarker as well as drug response related prediction.