Quick PRS is essentially an approximate version of MegaPRS. When running MegaPRS, the first step is to estimate Per-Predictor Heritabilities. In this step, it is necessary to calculate a tagging file and heritability matrix (the estimates of per-predictor heritabilities are then obtained by regressing the summary statistics on the tagging file and heritability matrix). The second step is to construct the prediction model. In this step, it is necessary to calculate predictor-predictor correlations, which are then used when estimating effect sizes.
Quick PRS speeds up the process by using pre-computed files (i.e., downloading pre-computed versions of the tagging file, heritability matrix, and predictor-predictor correlations). Although this is much faster, the final prediction model will likely be less accurate than the one constructed using the regular version of MegaPRS. This is because, ideally, the set of SNPs used to generate the pre-computed files should match those in the summary statistics (after quality control). However, in practice, we find the loss of accuracy is slight (less than a few percent), provided the summary statistics contain the majority of SNPs used to generate the pre-computed files (e.g., more than 90%).
