Ditte Demontis

Improved prediction of genetic predisposition to psychiatric disorders using genomic feature best linear unbiased prediction models

Publikation: KonferencebidragPosterForskning

Introduction: Accurate prediction of unobserved phenotypes from observed genotypes is essential for the success in predicting disease risk from genotypes. However, the performance is somewhat limited. Genomic feature best linear unbiased prediction (GFBLUP) models separate the total genomic variance into components capturing the variance by a genomic feature (e.g. GO term) and the remaining genomic variance by differential weighting of the genetic variants within the two groups. Previously we have demonstrated (on pigs and fruit flies) increased predictive ability when the genomic feature is enriched for causal variants. Here we apply the GFBLUP model to a small schizophrenia case-control study to test the promise of this model on psychiatric disorders, and hypothesize that the performance will be increased when applying the model to a larger ADHD case-control study if the genomic feature contains the causal variants.

Materials and Methods: The schizophrenia study consisted of 882 controls and 888 schizophrenia cases genotyped for 520,000 SNPs. The ADHD study contained 25,954 controls and 16,663 ADHD cases with 8,4 million imputed genotypes.

Results: The predictive ability for schizophrenia for the null model (all SNPs weighted equally) was low (0.07). Few GO terms did show a tendency of increased predictive ability; e.g. GO:0008645 had a predictive ability of 0.11 (unadjusted t-test p-value = 7.4x10-5), and explained 9% of the genomic variance, and 1.5% of the total phenotypic variance (0.6% for the null model).

Conclusion: The improvement in predictive ability for schizophrenia was marginal, however, greater improvement is expected for the larger ADHD data.
OriginalsprogEngelsk
Udgivelsesår27 maj 2017
StatusUdgivet - 27 maj 2017
BegivenhedESHG 2017: European Society of Human Genetics Annual Meeting - Copenhagen, Danmark
Varighed: 27 maj 201730 maj 2017

Konference

KonferenceESHG 2017: European Society of Human Genetics Annual Meeting
LandDanmark
ByCopenhagen
Periode27/05/201730/05/2017

Se relationer på Aarhus Universitet Citationsformater

ID: 113755226