Objective: To prioritize candidate genes from Parkinson’s disease (PD) genome-wide association studies (GWAS) for future functional studies. According to PubMed, most of the field still focuses on six to seven PD genes discovered decades ago.

Background: Ninety independent genetic variants within 78 loci were associated with PD GWAS.[1] Most GWAS variants are noncoding, where the causal gene is unclear.[2] There is a challenge to discover the biological function of the variants and nominate candidate genes. Previous studies leveraged machine learning models to predict causal genes from multi-omic data.[3] However, there is a lack of tissue-specific and cell-type relevant to PD, such as dopaminergic neurons. This study included genomic, transcriptomic and epigenetic datasets from dopaminergic neurons (DA) with other publicly available brain tissue to train a Parkinson’s disease-specific model.

Method: We trained a machine learning model using XGBoost[4] to score genes within the PD GWAS loci. XGBoost uses gradient boosting trees to perform classification. This model includes distance measures, expression quantitative trait loci (eQTL), single-cell RNA sequencing of PD DA neurons and enhancer-promoter interactions. We also performed fine-mapping of GWAS variants using echolocatoR[5] to select candidate variants. We labeled well-established PD genes such as GBA, GCH1, LRRK2, MAPT, SNCA, TMEM175, and VPS13C as positive and other genes that are one megabase pair upstream and downstream as negative. We performed posthoc analyses to support our findings, such as gene set enrichment analysis, pathway-specific polygenic risk scores (PRS), rare variant burden tests and differential gene expression.

Results: We identified candidate genes from the inositol phosphate biosynthetic process (GO:0032958), such as IP6K2, ITPKB, PPIP5K2, INPP5F and genes with missense variants such as SPNS1 and MLX. Pathway-specific PRS of inositol was associated with PD status. Differential gene expression nominated INPP5F, IP6K2 and MLX.

Conclusion: In this study, we nominated six candidate genes and one pathway for additional studies. A better understanding of the role of these genes in PD could lead to novel therapeutic targets for drug development.

References: 1. Nalls MA, Blauwendraat C, Vallerga CL, et al. Identification of novel risk loci, causal insights, and heritable risk for Parkinson’s disease: a meta-analysis of genome-wide association studies. The Lancet Neurology. 2019;18(12):1091-102.
2. Maurano MT, Humbert R, Rynes E, et al. Systematic Localization of Common Disease-Associated Variation in Regulatory DNA. Science. 2012;337(6099):1190-5.
3. Mountjoy E, Schmidt EM, Carmona M, et al. An open approach to systematically prioritize causal variants and genes at all published human GWAS trait-associated loci. Nature Genetics. 2021 2021/11/01;53(11):1527-33.
4. Schilder BM, Raj T. Fine-mapping of Parkinson’s disease susceptibility loci identifies putative causal variants. Human Molecular Genetics. 2021;31(6):888-900.

« Back to 2023 International Congress

MDS Abstracts - https://www.mdsabstracts.org/abstract/machine-learning-reveals-candidate-genes-from-parkinsons-disease-associated-loci/