Speaker Profile

Joydeep Sarkar

Biography

Joydeep Sarkar has a background in Biomedical Engineering, specialising in simulation modelling with over 10 years of experience in healthcare analytics. Prior to joining Holmusk, Joydeep worked in consulting with major providers and medical device & pharmaceutical companies, developing advanced analytics solutions. On the side he continues to work in healthcare research and has a proven track record of publications and presentations.

AI and Data Science Showcase: Holmusk
We build innovative, scalable and cost-effective digital behaviour change programs that combine cutting-edge clinical research, technology and design to guide people toward sustainable changes for better health. We develop powerful predictive algorithms that offer actionable insights for personalised care and population health management.

Towards a Platform For Personalized Medicine: Bridging The Gap Between Systems Pharmacology And Machine Learning With Real World Data
Before the advent of real world data (RWD) from electronic health records (EHRs), the gold standard was clinical trials. Quantitative systems pharmacology (QSP) has developed significantly over the last decade to use clinical trial data and published scientific information to develop simulation models that have significantly extended our understanding of diseases. However, such models generally are not equipped to capture the variability observed in large patient level longitudinal data now available from EHRs. On the other hand, data driven statistical models or the new generation of deep learning models applied to such EHR datasets can encompass the data but significantly lack in scientific context & mechanistic understanding. Holmusk has developed a hybrid modeling approach which takes its inspirations from QSP modeling but allows machine learning models to learn from the large RWD repositories. In doing so, the models keep the causal biological integrity but are readily applicable to longitudinal data from real patients with multiple comorbidities, polypharmacy, sparse and inconsistent data capture. Holmusk has also been working extensively to develop the process to ingest raw data and make it usable for models; create a large database of medications with relevant chemical properties & published efficacies. We will present multiple examples of how RWD from EHRs has been used by Holmusk to develop predictive models for use by clinicians and Pharma companies.