A lot has been written about the potential for
genomic information to revolutionize medicine. Much of the excitement centers
around the idea that if we know an individual’s genome we can use this
information to predict risk of a particular disease and then give specific
treatment. In a research article published
in the latest issue of Genome Medicine,
Cecile Janssens, Muin Khoury and colleagues improve current models of
genetic risk, taking us one step nearer to this aim.
Genetic risk prediction is a very active area of
research, boosted by the number of
genetic variants associated with common diseases that have been recently discovered
through genome-wide association studies. Previous studies looked at how single
mutations affect risk, but increasingly scientists and clinicians are finding
that multiple mutations contribute to disease and so risk prediction needs to
take into account multiple parameters.
The authors explore how several parameters of risk
prediction, especially sensitivity and positive predictive value (PPV) vary
under different models. They find that sensitivity and PPV are jointly maximized when the proportion of individuals identified at
high risk by the test equals the population disease frequency. Otherwise,
either the sensitivity is high or the PPV is high, but not both.
These results will help to advance strategies for predicting
disease risk based on genetic data, as we move towards an era of personalized