One might also anticipate that methodological developments in computing PRS may lead to better polygenic risk models and predictors. Competitors to the standard thresholding PRS, incorporating information about underlying linkage disequilibrium structure, are actively being developed (for example, LassoSum, LDpred, and PRS-CS) 44– 46. Such improvements may include the incorporation of information about the relationship of genotype to gene expression in the brain, a rapidly developing knowledge base, which would have beneficial effects in optimally weighting and calibrating the calculation of the PRS. In fact, the PsychENCODE Consortium recently showed that incorporation of information from an expression database may increase the predictive power of genotypic data by over threefold 47. One may expect a new generation of PRSs based on methods such as this.
