We improved the predictive power of several PRS for complex diseases using
the Discover module in our SaaS pipeline.
Our new scores were developed using the clinical datasets from across the
world and show comparisons with leading scores from the PGS Catalogue
applied to the same dataset.
Collated datasets of relevant genome-wide association statistics for each disease
Used Discover to assess 10 top PRS algorithms to find the best performing score on clinical data from multiple ancestries
Validated the best scores on an independent set of individuals from multiple datasets
Based performance stats on a model including genotyping array PCs 1-4 and using age and sex as covariates where relevant
Generated a set of performance statistics for both Allelica's PRS and a comparative PGS Catalog PRS for the same disease
Bolli et al 2021
Allelica
6.630.150
1.926.521
0.818 (0.811-0.824)
0.822 (0.815-0.829)
1.800 (1.773-1.827)
1.900 (1.872-1.978)
10.5%
21%
Allelica
313
577.113
0.679 (0.669-0.688)
0.708 (0.698-0.717)
1.557 (1.531-1.583)
1.807 (1.777-1.836)
1%
8%
Allelica
147
682.397
0.826 (0.815-0.837)
0.848 (0.837-0.858)
1.982 (1.927-2.035)
3.871 (3.711-4.043)
2%
12.5%
Allelica
6.917.436
620.612
0.626 (0.619-0.633)
0.632 (0.623-0.638)
1.587 (1.571-1.605)
1.620 (1.602-1.637)
1%
2%
Allelica
6.730.541
445.014
0.637 (0.623-0.652)
0.653 (0.640-0.667)
1.650 (1.615-1.680)
1.740 (1.704 - 1.770)
7%
11.5%
Allelica
186.726
247.151
0.645 (0.595-0.694)
0.718 (0.715-0.722)
1.412 (1.402-1.422)
1.552 (1.541 - 1.565)
-
-
