With the wide adoption of electronic health record systems, cross-institutional genomic medicine predictive modeling is becoming increasingly important, and have the potential to enable generalizable models to accelerate research and facilitate quality improvement initiatives. For example, understanding whether a particular variable has clinical significance depends on a variety of factors, one important one being statistically significant associations between the variant and clinical phenotypes. Multivariate models that predict predisposition to disease or outcomes after receiving certain therapeutic agents can help propel genomic medicine into mainstream clinical care. However, most existing privacy-preserving machine learning methods that have been used to build predictive models given clinical data are based on centralized architecture, which presents security and robustness vulnerabilities such as single-point-of-failure. In this proposal, we will develop novel methods for decentralized privacy-preserving genomic medicine predictive modeling, which can advance comparative effectiveness research, biomedical discovery, and patient-care.
Our first aim i s to develop a predictive modeling framework on private Blockchain networks.
This aim relies on the Blockchain technology and consensus protocols, as well as the online and batch machine learning algorithms, to provide an open-source Blockchain-based privacy-preserving predictive modeling library for further Blockchain-related studies and applications. We will characterize settings in which Blockchain technology offers advances over current technologies.
The second aim i s to develop a Blockchain-based privacy-preserving genomic medicine modeling architecture for real-world clinical data research networks.
These aims are devoted to the mission of the National Human Genome Research Institute (NHGRI) to develop biomedical technologies with application domain of genomics and healthcare. The NIH Pathway to Independence Award provides a great opportunity for the applicant to complement his computer science background with biomedical knowledge, and specialized training in machine learning and knowledge-based systems. It will also allow him to investigate new techniques to advance genomic and healthcare privacy protection. The success of the proposed project will help his long-term career goal of obtaining a faculty position at a biomedical informatics program at a major US research university and conduct independently funded research in the field of decentralized privacy-preserving computation.
The proposed research will develop practical methods to support privacy-preserving genomic and healthcare predictive modeling, and build innovations based on Blockchain technology for secure and robust machine learning training processes. The development of such privacy technology may increase public trust in research and quality improvement. The technology we propose will also contribute to the sharing of predictive models in ways that meet the needs of genomic research and healthcare.
| Kuo, Tsung-Ting; Kim, Hyeon-Eui; Ohno-Machado, Lucila (2017) Blockchain distributed ledger technologies for biomedical and health care applications. J Am Med Inform Assoc 24:1211-1220 |
