Develop and apply data science and AI technologies across all of NIH In recent decades, biomedical research has become an increasingly quantitative endeavor. Advances in data science and machine learning technologies now hold promise to acceler- ate this process, speeding even further the pace of discovery, development and application. Some estimates suggest that the amount of data generated in just a few years could soon surpass that collected in all prior human history. Individual researchers struggle to remain atop the flood of results in their narrow field. Clearly, discovery is being slowed and opportunities are being missed simply because researchers cannot possibly know, much less assimilate and relate to their studies, anything close to the full array of relevant information. Machine learning will greatly accelerate scientific discoveries. Machine learning (ML) technologies, including especially but not exclusively artificial intelligence (AI), offer powerful tools to aggregate, integrate, and perceive patterns across myriad data types, and predict structures, dynamics and interactions of molecules or populations, in applications that span every element of biomedical research, public health, and health care. AI technologies can increase efficiencies and reduce costs across the life cycle of drug development — target identification, molecular design and testing, clinical trials, manufacturing and post-marketing evaluation. AI tools can also dramatically facilitate basic, curiosity-driven research, on which subsequent development and health applications depend. For example, AI has predicted more than 200 million protein molecular structures that previously were painstakingly determined one-by-one by more costly and slower exper- imental procedures. The predicted structures, in turn, enable new proteins to be designed in silico to carry out specific functions. AI algorithms can surveille imaging data, e.g., X-rays, MRIs, CT scans, increasing the speed and accuracy of diagnostic and clinical decision- support. AI tools can integrate or fractionate population-level health data to identify or predict community health risks. Generative AI tools are being developed that create structured data from recorded doctor-patient interactions and components of patient electronic health records.
Made with FlippingBook - Online magazine maker