Using the optimal method-explained variance weighted genetic risk score to predict the efficacy of folic acid therapy to hyperhomocysteinemia

Background: Genetic risk score (GRS) is a useful way to explore genetic architectures and the relationships of complex diseases. Several studies had revealed many single nucleotide polymorphisms (SNPs) associated with the efficacy of folic acid treatment to hyperhomocysteinemia (HHcy).

Methods: We aimed to construct and screen out the optimal predictive model based on four GRSs and traditional risk factors. Four GRSs enrolled four SNPs (MTHFR rs1801131, MTHFR rs1801133, MTRR rs1801394, BHMT rs3733890) were presented as follows: (a) simple count genetic risk score (SC-GRS), (b) direct logistic regression genetic risk score (DL-GRS), (c) polygenic genetic risk score (PG-GRS), and (d) explained variance weighted genetic risk score (EV-GRS). We performed a prospective cohort study including 638 HHcy patients. Then we evaluated the associations of four GRSs with folic acid's efficacy and the performance of four GRSs.

Results: Four GRSs were independently associated with efficacy of treatment (p < 0.05). When combining GRSs with traditional risk factors, the AUC of the four models were all above 0.900 in the training set (Tradition + SC-GRS: 0.909, Tradition + DL-GRS: 0.909, Tradition + PG-GRS: 0.904, Tradition + EV-GRS: 0.910). And EV-GRS got the highest AUC. When evaluating the models in the testing set, we got the same conclusion that EV-GRS was optimal among four GRSs with the highest AUC (0.878) and the highest increase of AUC (0.008).

Conclusion: A more precise predictive model combing the optimal GRS with traditional risk factors was constructed to predict the efficacy of folic acid therapy to HHcy.