2024-10-22

091_ Midrange Filter – data application for post imputation quality control

Research Question and Aims

Genotype imputation is a widely used method to increase power and discoverability of signals in genome-wide association studies (GWAS), among other applications. However, significant SNPs in the midrange of imputation quality are not separable in terms of true associations, i.e. would be discovered in a fully sequenced data set, and false associations induced by the imputation process. For this reason, there is no clear recommendation for setting a threshold for imputation quality, since a high threshold might discard true associations while a low threshold might preserve false associations.
The Midrange Filter is a new method aiming to solve this problem by leveraging different output formats and aggregating SNPs into spikes, which is how association signals are interpreted in GWAS. The publication of the Midrange Filter is currently under review and an application on a real data set is part of the revision, where we would like to use the PsyCourse data set because of previous successful collaborations (PMID: 35232513). Please see the attached manuscript for details of the publication and the poster presenting the main findings at the IGES 2024 conference for a quick overview.

Analytic Plan

We plan to perform GWAS-style association analysis on the baseline data of the PsyCourse dataset and then perform quality control with the established imputation quality thresholds of 0.3 and 0.8 and contrast these results with the Midrange Filter. If time allows for more complex models, Wendel et. al. (PMID: 34247186) previously performed a longitudinal GWAS on the PsyCourse date set, which used the 0.8 threshold. Performing the same analysis with the Midrange filter might result in the discovery of more association signals, which were discarded by the threshold previously.

Resources needed

v1_id
v1_stat
v1_center
v1_sex
v1_age
v1_scid_dsm_dx_cat
v1_nrpsy_tmt_B_rt
v1_nrpsy_dgt_sp_bck
v2_nrpsy_tmt_B_rt
v2_nrpsy_dgt_sp_bck
v3_nrpsy_tmt_B_rt
v3_nrpsy_dgt_sp_bck
v4_nrpsy_tmt_B_rt
v4_nrpsy_dgt_sp_bck
psyc_id