Isaac Samuel Kohane, MD PhD, Chair of the Department of Biomedical Informatics

Overview

The COVID-19 pandemic poses a challenge for machine learning (ML) algorithms trained on pre-COVID data. Clinical presentations, disease trajectories, radiographic interpretations, and recommended medications for COVID all represent a significant departure from previous real-world practice patterns and may decrease the performance of established ML algorithms. 

In this webinar, Zak Kohane, MD, PhD, discusses this COVID discontinuity in the context of a longstanding issue in the field of AI and machine learning: how can algorithms (and the human physicians interacting with them) adapt to shifts in medical knowledge, whether from a novel therapy or a newly described disease? He outlines potential opportunities for AI researchers and steps that AI experts and clinicians should take to optimize system performance and patient care.

About the Presenter

Over the last 30 years, the research agenda of Isaac Kohane has been driven by the vision of what biomedical researchers could do to find new cures, provide new diagnoses, and deliver the best care available if data could be converted more rapidly to knowledge and knowledge to practice. In so doing, Kohane has designed and led multiple internationally adopted efforts to “instrument” the healthcare enterprise for discovery and to enable innovative decision-making tools to be applied to the point of care. At the same time, the new insights afforded by ’omic-scale molecular analyses have inspired him and his collaborators to work on re-characterizing and reclassifying diseases such as autism, rheumatoid arthritis, and cancers. In many of these studies, the developmental trajectories of thousands of genes have been a powerful tool in unraveling complex diseases. Kohane currently leads three NIH-funded projects that cut across the entire agenda:

  1. A Center for Excellence in Genomic Science to study neuropsychiatric disease at multiple levels (from molecular characterization of induced neurons obtained from fibroblasts of patients to automated classification of the textual component of their electronic medical record).
     
  2. The Coordinating Center for the Undiagnosed Diseases Network. Where patients with rare and unknown diseases are provided with combined clinical and molecular diagnoses in a nationally-scaled infrastructure so that they can see the right expert with all their relevant data at hand.
     
  3. A Center for Excellence in Big Data to Knowledge to both create a nationally scaleable research data-sharing infrastructure and demonstrate its use for neurodevelopmental diseases.
     

Kohane is always on the look out for like-minded “quants” who share the same goals to bring a better future for medicine and biomedical science to the present.