Publication Spotlight


Polygenic Risk Scores (PRS) are predictive models of disease risk that are constructed by combining the estimated effects of genetic variants across the genome. While PRS do seem to have promise when designed and vetted correctly, many critical questions are still open: For which diseases and traits can PRS improve upon clinical predictive models? How predictive is a given PRS for a specific scored quantile and what is the right way to use this predictive ability in a clinical setting? How robust and generalizable are PRS scores when applied to populations not used to construct the PRS?

Here we're spotlighting a paper now in review by Scott Kulm (PhD candidate at Weill Cornell Medicine, co-advised by Dr. Mezey and Dr. Olivier Elemento) describing an exhaustive investigation of the accuracy of PRS for 23 diseases where data is available in the UK Biobank, where the objective is to answer components of these open PRS questions:

A Systematic Framework for Assessing the Clinical Impact of Polygenic Risk Scores

Scott Kulm, Andrew Marderstein, Jason Mezey and Olivier Elemento

medRxiv / Synopsis@KulmScott

2021 Kulm figs in review

Above are two subfigures from the paper summarizing prediction performance and the predictive value of PRS: the left showing the increase in the area under the Receiver Operating Characteristic curve (AUC) when adding a PRS to a base clinical prediction model and the right showing the proportion of individuals in the top 5% of disease risk according to the base clinical model that would have lower risk when adding a PRS (see a version of the paper available on medRxiv for further details). Among the main conclusions described in the paper:

  • For certain diseases, there is a high false positive rate for clinical assessments of "high risk" where an appropriately constructed PRS can bring this false positive rate down


  • Constructed appropriately, a PRS can be relatively independent of clinical and life-style factors, such that a PRS can add to prediction performance in a straightforward way 


  • Individuals with a high PRS benefit significantly more from lifestyle modification compared to individuals with a low PRS

If requesting additional information or if you have questions for the lead author please email us at MezeyLab