Estimating Polygenic Risk Scores (PRS) on longitudinal changes in Imaging Derived Phenotypes (IDPs)
Background
The deep phenotyping of populations is increasingly being used to investigate how the brain, genetics, behavior, and clinical outcomes interact. Through the collaborative collection of large samples, like the UK Biobank (UKBB), the methods and interpretation of these kinds of datasets can be reported and applied to real world outcomes. Large population samples can be used for normative modelling to contrast a clinical sample, like the Quebec Parkinson’s Network (QPN) or the Parkinson’s Progression Markers Initiative (PPMI).
In order to maximize the information available within these samples, it is important to combine different modalities of information, like neuroimaging and genetics.
A common way to combine genetics and neuroimaging data is with genome-wide association studies (GWAS) and the estimation of polygenic risk scores (PRS). A PRS model can be estimated for each neuroimaging feature, or an imaging derived phenotype (IDP). This provides a model that highlights the single nucleotide polymorphisms (SNPs) that contribute the most variation to the IDP, linking the genetic contribution to neural expression.
Methods
My work revolves around contributing high quality imaging derived phenotypes (IDPs) across multiple datasets (UKBB, QPN, PPMI). This involves utilizing current standard pre-processing tools for MRI data, like fMRIPrep and TractoFlow. The analysis, quality control, and indexing of these derivatives is being contributed through other ongoing projects in the lab, namely Nipoppy and NeuroBagel.
With these high-quality IDPs generated, I am estimating the variational inference of polygenic risk scores (VIPRS). I am extending this approach by estimating the PRS of the longitudinal change of the IDPs from the samples. This will more accurately target the SNPs that correspond to change over time in the IDPs. This will improve our ability to use these biobanks to identify the neural and genetic features that correspond with degenerative diseases.
Goal
My goal with creating these higher quality imaging derivatives (IDPs) is to combine them with other physiological phenotypes in the sample (genetics, behavior, clinical status, etc.). Incorporating polygenic risk scores (PRS) with the imaging phenotypes will allow researchers to maximize the use of information in these biobanks. By extending this usage to incorporate longitudinal change in the PRS estimates of the IDPs we can better identify the multi-modal features that underlie the transition toward a disease.
People
