Skip to main content
NCBI home page
Wiley Open Access Collection logoLink to Wiley Open Access Collection
. 2026 Jun 27;50(6):e70045. doi: 10.1002/gepi.70045

Individualized Bayesian Inference Identifies Novel Genetic Variants for Parkinson's Disease

Jin Ren 1, Yasaman J Soofi 2, Md Asad Rahman 1,2, Qing Lu 3, Jinling Liu 1,✉
PMCID: PMC13428404  PMID: 42363869

ABSTRACT

Parkinson's disease (PD) is a complex neurodegenerative disorder with a significant genetic component. While genome‐wide association studies (GWAS) have been instrumental in identifying genetic variants associated with PD, the reliance on large sample sizes and population‐level analyses may overlook variants with lower minor allele frequencies or individual‐specific relevance. Individualized Bayesian Inference (IBI) offers a promising method to complement GWAS by identifying and prioritizing candidate genetic markers at both the individual and patients‐like‐me subgroup levels. This study evaluates the application of IBI to PD genetics, using GWAS as a baseline for comparison. We analyzed genetic data from the Fox Insight online study, including 8840 individuals (8585 PD cases and 255 controls). IBI prioritized variants that were not detected or were ranked substantially lower by GWAS, including variants within or near genes with prior PD association. The top 200 IBI SNPs showed stronger predictive performance in ANN models (AUC = 0.79) than the top 200 GWAS SNPs (AUC = 0.72), providing complementary support for the utility of IBI‐based prioritization in this cohort. Notably, IBI highlighted variants with lower minor allele frequencies that GWAS did not detect. This study demonstrates the utility of IBI as a complementary tool for prioritizing PD‐related candidate variants and genes for further investigation.

1. Introduction

Parkinson's disease (PD) is a common neurodegenerative disease (NDD) characterized by the progressive loss of dopaminergic neurons in the substantia nigra, resulting in various motor and non‐motor symptoms (Dauer and Przedborski 2003; La Spada and Ranum 2010). Motor symptoms of PD include tremors in the hands, arms, legs, jaw, or head, muscular rigidity, slowness of movement, and impaired balance and coordination, which can lead to falls (Jankovic et al. 1999; Nutt and Wooten 2005; Jankovic 2008; Wilczyński et al. 2017). Additionally, individuals with PD often experience a wide range of non‐motor symptoms, including mood disturbances, cognitive changes, sleep disturbances, and autonomic dysfunction, which significantly impact their overall quality of life (Doty et al. 1984; Gagnon et al. 2006; Goldstein 2006; Chaudhuri and Schapira 2009; Aarsland et al. 2017). With over 10 million individuals worldwide suffering from PD and approximately 90,000 new diagnoses in the United States each year (Dorsey et al. 2018; Marras et al. 2018; Parkinson's Foundation 2023), it is crucial to understand its underlying mechanisms and develop more effective treatments.

While the molecular mechanism underlying PD is still not fully understood, genetic factors play a significant role in its development. The discovery of key genes such as SNCA, LRRK2, and PARK7 has provided critical insights into both familial and sporadic forms of the disease, linking genetic mutations with pathophysiological mechanisms of PD (Polymeropoulos et al. 1997; Satake et al. 2009; Nuytemans et al. 2010; Lesage et al. 2010; Nalls et al. 2014). Furthermore, genes related to mitochondrial function and cellular stress responses, such as PINK1 and PARK2 (Nuytemans et al. 2010; Narendra and Youle 2011; Pilsl and Winklhofer 2012), have been shown to be associated with PD, suggesting the involvement of mitochondrial dysfunction in the disease (Sandyk 1993; Cereda et al. 2011; Xu et al. 2011). These findings highlight the complex and multifactorial nature of PD, prompting the use of Genome‐Wide Association Studies (GWAS) to identify loci linked to PD and providing insight into its genetic architecture(Valente et al. 2004; Mata 2004; Singleton and Hardy 2019; Lubbe et al. 2021; Bryant et al. 2021; Rahman and Liu 2023).

Despite the successes of GWAS in identifying significant genetic risk factors and suggesting novel therapeutic targets, the often reliance on large sample sizes and the focus on population‐level associations can miss variants with lower minor allele frequency (MAF) that may be important for disease pathogenesis (Chang et al. 2017; Nalls et al. 2019). This limitation has highlighted the need for alternative methods that can identify individual‐ or subgroup‐specific genetic variants missed by traditional approaches. Individualized Bayesian Inference (IBI) offers a promising method for identifying such variants by focusing on individual‐level genetic data. IBI has previously demonstrated its potential in identifying single‐nucleotide polymorphisms (SNPs) related to hypertension (Rahman et al. 2023). In that study, IBI was able to detect low‐MAF variants missed by GWAS in the same cohort, providing a more personalized approach to genetic analysis. This study seeks to apply IBI to PD genetics, using GWAS as a baseline method for comparison. The primary goal is to evaluate whether IBI can prioritize PD‐related candidate variants, including low‐MAF variants that may be under‐ranked by population‐level association analysis.

In our study, GWAS was performed without adjusting for covariates such as age, sex, or principal components to maintain consistency with IBI's framework, which does not yet incorporate these covariates. While our previous studies (Rahman and Liu 2023), have shown that adjusting for covariates provides valuable insights, the choice to omit covariates in our current GWAS analysis ensures a comparable population‐level association baseline with IBI. Notably, the GWAS results without covariates show significant overlap with our prior study that adjusted for covariates (Rahman and Liu 2023). Additionally, we did not perform linkage disequilibrium (LD) pruning because the goal was not to identify independent loci, but to compare SNP prioritization between IBI and GWAS. We also chose not to impute the data, focusing instead on using the SNPs available within the data array to investigate IBI's potential without the need for imputation‐based methods.

2. Materials and Methods

2.1. Genome‐Wide Association Studies

GWAS are commonly used to identify genetic variants associated with specific traits or diseases. In GWAS, the genetic variations across the entire genome of a large group of individuals with and without the trait or disease of interest are analyzed. In our study, we applied Fisher's exact test to the entire population (denoted as D) to calculate the p‐value of each genetic variant. To ensure robustness and minimize false positives caused by multiple hypothesis testing, we adopted the Bonferroni correction (Korte and Farlow 2013).

In our previous study, we have conducted a comprehensive GWAS analysis on this same dataset by adjusting covariates such as age, sex, and principal components, as well as looked into independent risk loci by performing LD pruning. In this current study, we focus to identify novel PD variants/genes using IBI, with GWAS as a baseline population‐level analysis. Since IBI does not adjust the non‐genomic factors yet, we used Fisher's exact test without covariate adjustment for GWAS analysis to ensure a fair comparison to IBI. Yet, the substantial overlap between the GWAS results with and without covariates (9 out of the top 10, and 17 out of the top 20) supports the robustness of the major signals and the main ranking patterns.

2.2. The Individualized Bayesian Inference (IBI) Method

The IBI algorithm developed by Rahman et al. employs Bayesian networks not only to identify rare and low‐MAF genomic variants but also to detect common variants that are specific to an individual (Rahman et al. 2023).

IBI uses a bipartite Bayesian network to model the probabilistic relationships between genomic variants (SNPs) and traits (e.g., PD status). The network consists of two sets of nodes: one representing the genomic variants and the other representing the traits. Directed edges between these nodes represent the conditional dependencies or predictive relationships from variants to traits.

To identify significant variants for an individual, IBI partitions the overall population into two subpopulations based on the presence or absence of a specific variant. For a given individual with a minor allele at a specific SNP, IBI creates a “patient‐like‐me” subpopulation consisting of all individuals who also have the minor allele at that SNP. The remaining individuals form the second subpopulation.

IBI then evaluates the influence of the SNP within the “patient‐like‐me” subpopulation by calculating the marginal likelihood of the SNP estimating the trait. This is done using the Bayesian Dirichlet equivalent uniform (BDeu) score. The same process is repeated for the remaining subpopulations using the best predictor SNP for that group. The overall marginal likelihood for the SNP is the sum of the marginal likelihoods from both subpopulations.

The IBI algorithm assigns a posterior probability to each SNP for its influence on the trait, considering the individual's specific genomic background. This individualized approach allows IBI to prioritize significant variants that may be missed by population‐level methods like GWAS, particularly those with low MAF.

In summary, IBI provides a personalized analysis by focusing on the genomic variants observed in an individual's genome and their specific subpopulation, making it a valuable complement to traditional GWAS in uncovering the genetic basis of complex traits like PD. For those interested in a more detailed explanation of the IBI method, please refer to the paper by Rahman et al. (2023).

2.3. Data Preprocessing

Genotyping of participants in Fox Insight was conducted on the V3, V4, and V5 platforms. The V5 platform, currently in use, employs a customized Illumina Infinium Global Screening Array with approximately 690,000 SNPs and was used for genotyping in 80.4% of the participants. To ensure data quality, additional filtration, and quality control measures were applied using PLINK with specific thresholds for missingness per individual (Mind ≤ 0.05), missingness per marker (Geno ≤ 0.03), minimum allele frequency (MAF ≥ 0.01), and Hardy–Weinberg equilibrium (HWE p‐value ≤ 10e‐6) on each chromosome. Finally, the resulting SNP values were transformed into 0 s and 1 s using dominant coding. The dataset consists of 447,089 SNPs and 8840 samples (Figure 1), with phenotype data based on the current PD diagnosis status. These variables were reported by participants who confirmed being diagnosed with PD by a healthcare professional.

FIGURE 1.

FIGURE 1

Open in a new tab

The integrative analysis framework. This schematic illustrates our integrative study's methodology, beginning with preprocessing approximately 690k single‐nucleotide polymorphisms (SNPs) and 11k individuals to a final dataset of 447,089 SNPs and 8840 individuals. The dataset is then split for training and testing, and the training set is employed in GWAS and Individualized Bayesian Inference (IBI) methods of statistical association analysis. Genome‐wide association studies (GWAS) are conducted on the entire population (denoted as D) to calculate the p‐value, whereas IBI is applied to sub‐populations (D Vs=1 and D Vs=0 ) to get the log marginal likelihood (M s,r ). The top 200 SNPs from both methods along with a set of randomly selected SNPs are selected for developing ANN prediction models. The final step involves a rigorous integrative analysis of high‐ranked variants using evaluation metrics, SHapley Additive exPlanations (SHAP) analysis, and a literature search to discover novel significant genetic markers as potential contributors to the genetic landscape of PD.

The processed dataset was then divided into a training set (80%, n = 7072 individuals) and a testing set (20% or 1768 individuals) while conserving the case‐control ratio. After the data splitting process (Figure 1), the test set remains concealed from the association search methods (GWAS and IBI) and model development to ensure the prevention of information leakage. GWAS and IBI were applied to the training set to examine the association between SNPs and PD status. The top SNPs from each method were obtained and used as the input features for the prediction model. Five sets of randomly selected SNPs were also used for benchmarking purposes. Prediction analysis as secondary validation. The prediction experiment was performed to provide additional evidence that IBI‐selected SNPs capture meaningful signals beyond those identified by GWAS. A single fixed discovery‐validation split was used, consistent with conventional GWAS validation frameworks where replication is evaluated in a predefined independent dataset. This approach allows for clear interpretability of the SNP sets and avoids the variability in selected variants that would result from repeated resampling.

2.4. Supporting Machine‐Learning Prediction Analysis and SHAP‐Based Prioritization

To provide complementary support for the variants prioritized by IBI, we performed a supporting machine‐learning analysis to compare the predictive signal of SNP sets selected by IBI with those selected by GWAS. Specifically, SNP sets ranked by IBI and GWAS were evaluated using three classifiers, artificial neural network (ANN), support vector machine (SVM), and random forest (RF), at multiple thresholds (top 50, 200, and 300 SNPs). Model performance was summarized primarily by the area under the receiver operating characteristic curve (AUC) on an independent test set.

Because the dataset is highly imbalanced (~97% PD cases and ~3% controls), class weights inversely proportional to class frequencies were applied during model training to partially account for class imbalance. These analyses were not intended to establish an optimized prediction framework, but rather to compare whether SNPs prioritized by IBI carry stronger relative signal than those prioritized by GWAS. Additional implementation details and hyperparameter settings are provided in the Supplementary Methods.

To further prioritize influential variants within the selected SNP sets, we applied SHapley Additive exPlanations (SHAP) analysis using the ANN models. Although SHAP values are derived from the prediction models, they were used here only as a complementary prioritization tool rather than as evidence of optimized predictive performance. SHAP values were computed on the independent test set using the SHAP Python library. For interpretability, we focused on the top 15 SNPs ranked by maximum absolute SHAP values. Maximum rather than mean absolute SHAP value was used to highlight SNPs with strong effects in a subset of individuals. The top 15 SNPs were selected because their SHAP values were markedly higher than those of the remaining variants, making them the most informative candidates for detailed biological interpretation. SHAP‐based prioritization was interpreted jointly with IBI ranking, GWAS ranking, and literature evidence. Variants highlighted by SHAP were therefore not treated as stand‐alone discoveries, but as candidate loci supported by multiple lines of evidence.

3. Results and Discussion

3.1. Discovery of Significant SNPs Using GWAS and IBI

The detailed workflow of our integrative analysis is shown in Figure 1, which began with the preparation of a dataset that had 11,000 people and about 690,000 SNPs. After processing, 447,089 SNPs from 8840 individuals were found to be appropriate for further investigation. We divided this dataset into training and testing sets, where the training set was utilized to apply both GWAS methods on the entire dataset (D) and IBI on specified sub‐populations (D Vs=1 and D Vs=0 ). These methods were employed to extract statistical associations and compute respective p‐values and log marginal likelihoods (M s,r ). The top 200 SNPs identified by each method and a control group of 200 randomly selected SNPs were then used to develop ANN prediction models. The process concluded with a thorough integrative analysis that included evaluation metrics, a thorough search of the literature, and SHAP analysis in order to find and validate novel significant genetic markers that could advance our knowledge of the genetic foundations of Parkinson's disease. The results obtained from our IBI analysis provide valuable insights into potential genetic variants that may be associated with PD, with GWAS serving as a baseline for comparison. Further validation is needed to establish the functional relevance of these variants.

The QQ plot (Figure 2a) generated from our GWAS experiment reveals a significant deviation from the expected values, indicating a strong association between the top‐ranked SNPs and PD. To effectively control for multiple testing in genetic association studies, we employed the Bonferroni correction, a widely accepted method for maintaining statistical significance. By dividing the conventional alpha level (0.05) by the number of tested variants, we applied the Bonferroni threshold of p‐value 1.1e‐7. Subsequently, we identified 21 GWAS SNPs that exceeded this significance threshold. Remarkably, all of the identified SNPs were located on chromosome 1, which was further graphically demonstrated by the regional association plot (Figure 2c). The plot illustrates the clustering of top‐ranked GWAS SNPs around position 1.55e8 on chromosome 1, indicating a region that warrants further investigation. We have highlighted the top 5 ranking SNPs and provided the names of the genes in which they are located. This highlighting serves to draw attention to the most promising candidates for PD.

FIGURE 2.

FIGURE 2

Open in a new tab

Genome‐wide association studies (GWAS) and Individualized Bayesian Inference (IBI) single‐nucleotide polymorphism (SNP) Analysis. (a) GWAS QQ plot with a Bonferroni‐adjusted threshold (dashed line) of 1.1e‐7, surpassed by 21 SNPs. (b) Comparison of MAF distribution between random SNPs and top SNPs from GWAS and IBI analyses. (c) Regional Association plot of GWAS p‐values for chromosome 1, showcasing a clustering of significant SNPs (dashed line indicates Bonferroni threshold). (d) Manhattan plot for IBI log marginal likelihood across all chromosomes, with a dashed line denoting the threshold for the top 21 SNPs (threshold value: 1.46).

In a comparison of MAF between the top 200 SNPs identified by GWAS and IBI, we observed that IBI tends to identify variants with much lower MAF, which are potentially missed by GWAS (Figure 2b).

This finding suggests that IBI may be particularly effective in detecting variants with a lower frequency that could contribute to PD susceptibility. However, the functional relevance of these variants requires further investigation. Overall, our results indicate that the PD's genetic architecture involves both common and low‐frequency variants, and combining the results from both methods may provide a more comprehensive understanding of the genetic basis of PD. Further investigation of the functional effects of the identified SNPs may provide insight into the underlying molecular mechanisms of PD pathogenesis and lead to potential therapeutic targets for the disease.

The IBI method offers a distinct approach to identifying PD‐associated SNPs that may not be captured by GWAS. In our study, IBI calculated the log marginal likelihood for each SNP, and the resulting Manhattan plot showed the distribution of −Log10 (normalized log marginal likelihood) across the chromosomes (Figure 2d). For descriptive comparison with the 21 Bonferroni‐significant GWAS SNPs, we examined the top 21 IBI‐ranked SNPs using a normalized log marginal likelihood threshold of 1.46. These SNPs were distributed across various chromosomes, unlike the GWAS results where all 21 top SNPs were located on chromosome 1. The top 5 ranking SNPs are shown alongside their respective gene locations. Interestingly, the top 10 SNPs identified by IBI, as shown in Table 1, provide key insights into PD genetics. While GWAS has identified several SNPs associated with PD, IBI reveals additional variants that might have been overlooked in traditional approaches (Rahman and Liu 2023).

TABLE 1.

Top 10 SNPs identified by the IBI method, with associated genes and their relevance to PD.

Variant ID MAF IBI rank GWAS rank Overlapped or nearest genes Association
i4000415 0.02 1 1 GBA PD (Usenko et al. 2021)
rs9933717 0.02 2 113 CDH8 ASD, HD, PD (Liu et al. 2011; Friedman et al. 2015)
rs6755210 0.04 3 213651 MIR4431 Metabolic diseases (Pan et al. 2022; Cai et al. 2023)
rs117887535 0.01 4 725 ANO2 MS (Ayoglu et al. 2016)
rs76630176 0.02 5 47297 LRP12; ZFPM2 Parkinsonism‐related, ALS, OPDM1 (Greenbaum et al. 2012; Greenbaum and Lerer 2015; Shimizu et al. 2022; Kume et al. 2023; Eura et al. 2024; Henden et al. 2024)
rs6942437 0.02 6 43573 CNTNAP2 PD(Infante et al. 2015; Murugesan and Rajagopal 2022)
rs77508540 0.01 7 35990 PLAC9P1 None
rs112941894 0.05 8 143900 DUSP10 Parkinsonism‐related, AD, ALS (Sanchez‐Contreras et al. 2018; Lai et al. 2021)
rs117235806 0.02 9 297075 PTPRT PD, AD, ALS (Liu et al. 2021; Lu et al. 2021)
rs138195872 0.01 10 353682 PTPRT PD, AD, ALS (Liu et al. 2021; Lu et al. 2021)
Open in a new tab

Note: The “Association” column summarizes prior literature relevant to the overlapped or nearest gene, including direct PD evidence where available and indirect evidence from related neurological or biological contexts.

Abbreviations: GWAS, genome‐wide association studies; IBI, Individualized Bayesian Inference; SNPs, single‐nucleotide polymorphisms.

Among the top 10 IBI SNPs, several are located in or near genes with established PD association, prior PD‐related evidence, or indirect neurologic relevance, supporting their biological plausibility as candidate loci, although further research is required to confirm these associations. Specifically, the top‐ranked SNP i4000415 (also known as rs76763715) has previously been reported to be associated with PD and is an established PD locus. In contrast, SNP rs9933717 (#2), while not directly associated with PD, is located near the gene CDH8. CDH8 has been associated with pathways crucial for high‐order executive functions and cognitive deficits seen in ASDs, as well as NDDs like PD and HD (Liu et al. 2011; Friedman et al. 2015). Similarly, rs6755210 (#3) is located near MIR4431, which is associated with Metabolic Syndrome (Pan et al. 2022; Cai et al. 2023). Considering the established connections between PD and Metabolic diseases, including diabetes (Sandyk 1993; Cereda et al. 2011; Xu et al. 2011; Pilsl and Winklhofer 2012), this locus has indirect PD relevance and may warrant further investigation.

Furthermore, rs117887535, ranked as #4 by IBI, is located in the ANO2 gene that is known to be associated with Multiple Sclerosis (MS) (Ayoglu et al. 2016). This provides indirect neurologic relevance, although prior PD‐specific evidence remains limited. SNP rs76630176 with an IBI rank of #5 is an intergenic variant located near LRP12 and ZFPM2. Gene ZFPM2 has been highlighted for its involvement in antipsychotic‐induced Parkinsonism and may offer insight to PD pathology (Greenbaum et al. 2012; Greenbaum and Lerer 2015). On the other hand, LRP12 is known to be related to NDDs such as Oculopharyngodistal Myopathy 1 (OPDM1) (Shimizu et al. 2022; Eura et al. 2024) and Amyotrophic Lateral Sclerosis (Kume et al. 2023; Henden et al. 2024). This variant's presence in this locus indicates a potential multi‐gene influence on neurodegeneration that makes it an interesting candidate for PD, although direct PD evidence remains limited. Similarly, the CNTNAP2 gene has previously been linked to neuronal development, synaptic function, and has prior PD‐related evidence (Infante et al. 2015; Murugesan and Rajagopal 2022). The 6th IBI variant rs6942437 would also be a potential PD candidate since it's located in CNTNAP2.

Additional gene loci like DUSP10 and PTPRT are also identified via the top 10 SNPs listed in Table 1. DUSP10, highlighted by the variant rs112941894 (#8), has broader neurodegenerative or Parkinsonism‐related relevance, whereas PTPRT, highlighted by variants rs117235806 and rs138195872 (ranked #9 and #10), has prior PD‐related evidence together with broader neurodegenerative relevance (Sanchez‐Contreras et al. 2018; Lai et al. 2021; Liu et al. 2021; Lu et al. 2021). The fact that 5 out of the top 10 IBI‐ranked SNPs that are situated in PD‐associated genes, but are ranked much lower by GWAS, highlights the potential of IBI in revealing additional and meaningful genetic insights in PD research.

3.2. PD‐Relevant Candidate Genes Prioritized by IBI but Under‐Ranked by GWAS

Our initial analysis focused on the top 10 variants ranked by IBI to investigate their potential contribution to PD, where 5 out of 10 variants with prior evidence on PD association were overlooked by GWAS. Expanding this scope, we examined the top 200 IBI‐ranked variants. This extended review allows for a deeper exploration into the genetic landscape of PD, moving beyond the constraints of GWAS, which primarily identifies common, rather than low‐frequency PD variants due to limited power. A prior study by Rahman and Liu (2023) conducted a comprehensive GWAS with adjustments for covariates and principal components, identifying significant PD‐associated variants in genes such as TRIM46, ASH1L, ARHGEF2, RIT1, PBXIP1, KCNN3, and LMNA, all located on chromosome 1. However, this approach still does not capture the full spectrum of genetic influences.

IBI addresses this gap by identifying SNPs that were overlooked by GWAS. Table 2 presents SNPs within PD‐associated genes, ranked within the top 200 by IBI but positioned far lower in GWAS rankings. The column “Gene's Best GWAS Rank” indicates the highest GWAS rank for any SNP located within or near the listed genes, underscoring the considerable disparity in rankings between the two methods. Interestingly, all SNPs listed in Table 2 have an MAF of less than 0.05, emphasizing their low frequency and potential significance in PD pathology. Without further performing this complementary IBI analysis, these PD‐associated SNPs or genes would have been missed with a GWAS‐only approach. Unlike the findings from Rahman et al. where all significant genes were located on chromosome 1, the genes listed in Table 2 are distributed across various chromosomes, illustrating the broad genetic diversity associated with PD.

TABLE 2.

Expanded analysis of top 200 IBI‐ranked variants in PD‐associated genes overlooked by GWAS.

Chr Variant ID MAF IBI Rank GWAS Rank Overlapped or nearest genes Gene's best GWAS rank
6 rs35727307 0.047 24 440,152 RPS6KA2 927
12 rs34637584 0.010 67 61,686 LRRK2 17,797
4 rs17020293 0.021 83 365,279 GRID2 1540
2 rs72771309 0.022 88 210,255 KCNS3 1531
8 rs4375045 0.014 94 219,272 ARHGEF10 8338
8 rs6989286 0.015 183 410,085 ARHGEF10 8338
8 rs11986753 0.015 190 410,119 ARHGEF10 8338
Open in a new tab

Abbreviations: GWAS, genome‐wide association studies; IBI, Individualized Bayesian Inference.

One noteworthy example is the variant rs35727307, located in RPS6KA2, a kinase gene involved in neuronal system development and survival. Its role in PD pathology has been suggested through studies highlighting its influence on neurodegeneration (Mortezaei et al. 2017; Fernández‐Santiago et al. 2019; Martín‐Flores et al. 2019). Similarly, gene LRRK2 has long been known as a region of interest for PD genomics (Gilks et al. 2005; Mata et al. 2006; Martin et al. 2014). Subsequent studies suggested that rs34637584 (a synonymous exonic variant and the most frequent mutation in LRRK2) is a potential explanatory factor for PD cases, suggesting its significant role in the disease's development (Singleton et al. 2013; Alessi et al. 1979).

IBI's ability to highlight such variants demonstrates its strength in pinpointing critical genetic factors. Moreover, IBI identified variants in GRID2, a gene crucial in synaptic function and dopaminergic neurons (Seo et al. 2020; Huang et al. 2022), both key components in PD pathophysiology. Similarly, IBI identified rs72771309 near KCNS3, while previous studies have reported an association between KCNS3 and PD risk(Hendrickx and Glaab 2020; Zheng et al. 2021). KCNS3 encodes a voltage‐gated potassium channel that is involved in regulating neuronal excitability (García‐Campayo et al. 2018; Hendrickx and Glaab 2020).

Among the other notable findings, rs4375045, rs6989286, and rs11986753 ranked high by IBI are located in ARHGEF10, a gene involved in various neuropsychiatric disorders (schizophrenia, autism, bipolar disorder, and depression), and NDDs (PD, and AD)(Glaab and Schneider 2015; Wüllner et al. 2016; García‐Campayo et al. 2018). According to a recent study, ARHGEF10 is a member of the Rho GTPase signaling molecule family, and disruptions with this pathway can cause anomalies in the nervous system (Bloch‐Gallego and Anderson 2023). These results suggest that IBI may have identified potential genetic candidates for PD risk that were not detected by GWAS, further highlighting its potential to contribute to understanding PD's genetic complexity.

3.3. Supporting Machine‐Learning Analyses for Signal Validation and SNP Prioritization

To provide complementary validation of the signal identified by IBI, we compared the predictive performance of SNP sets prioritized by IBI and GWAS using ANN, SVM, and RF classifiers across multiple selection thresholds (top 50, 200, and 300 SNPs). Model performance was evaluated on an independent test set using AUC (Table 3).

TABLE 3.

Predictive performance (AUC) of three classifiers, ANN, SVM, and RF.

Method Top k variants AUC
ANN SVM RF
GWAS 50 0.7 0.7 0.7
200 0.72 0.69 0.7
300 0.66 0.68 0.64
IBI 50 0.75 0.73 0.74
200 0.79 0.77 0.74
300 0.78 0.77 0.74
Open in a new tab

Across all classifiers and thresholds, SNPs selected by IBI consistently demonstrated superior predictive performance compared to those selected by GWAS (Table 3). For example, at the top‐200 threshold, IBI achieved AUCs of 0.79, 0.77, and 0.74 with ANN, SVM, and RF, respectively, compared to 0.72, 0.69, and 0.70 for GWAS. Similar trends were observed at the top‐50 and top‐300 thresholds, indicating that IBI prioritizes SNPs with stronger relative signal.

As a baseline comparison, models trained on randomly selected SNP sets showed no predictive discrimination (AUC ≈ 0.51), confirming that both GWAS‐ and IBI‐selected SNPs capture non‐random genetic information. Additional performance metrics, ROC, and learning curves are provided in the Supplementary Materials (Supplementary Table S1 and Supplementary Figure S1). Given the severe class imbalance of the dataset, absolute prediction metrics should be interpreted with caution. Accordingly, these analyses are presented as supporting validation, and predictive performance is interpreted comparatively rather than as evidence of a fully optimized prediction framework, with emphasis on the consistently higher AUC achieved by IBI‐prioritized SNPs than by GWAS‐prioritized SNPs.

To further prioritize influential variants, SHAP analysis was applied to the ANN models. The SHAP plots highlighted SNPs with strong contributions to model predictions within both GWAS‐ and IBI‐prioritized SNP sets (Figure 3).

FIGURE 3.

FIGURE 3

Open in a new tab

SHAP Plots for genome‐wide association studies (GWAS) and Individualized Bayesian Inference (IBI) single‐nucleotide polymorphism (SNP) Analyses. The top 15 SNPs, ranked by maximum absolute SHAP values, are depicted in the SHAP plots for (a) GWAS SNPs and (b) IBI SNPs. Each row represents a single SNP, and each dot corresponds to an individual in the test dataset. Dot colors indicate genotype status under dominant coding: red denotes the presence of the SNP (at least one minor allele), while blue denotes its absence (two major alleles). The horizontal position reflects the SHAP value, which quantifies the contribution of each SNP to the model output. Positive SHAP values indicate an increased likelihood of Parkinson's disease prediction, whereas negative SHAP values indicate a decreased likelihood. These plots highlight influential SNPs and support their prioritization for further biological investigation.

Four SNPs, i4000415, rs9933717, rs61817862, and rs77140267, were notably present in the SHAP plots of both GWAS and IBI analyses. The variants i4000415 and rs9933717 were already known to have an association with PD (Liu et al. 2011; Usenko et al. 2021). The first SNP in both plots is i4000415, also known as rs76763715. It holds the first rank in both IBI and GWAS analyses. The SNP i4000415 is located in the GBA gene, a locus already well‐recognized for its association with PD (Usenko et al. 2021). The GBA gene is known to hold several PD‐related SNPs, and our findings further reinforce its significance. Furthermore, the second variant appearing in both SHAP plots is rs9933717, which is located in the CDH8 gene. This gene has been implicated in pathways crucial for high‐level executive functions, which are linked to the cognitive impairments observed in autism spectrum disorders (ASDs), as well as in NDD like Parkinson's and Huntington's disease (HD) (Liu et al. 2011; Friedman et al. 2015). The consistent appearance of rs9933717 in the SHAP analyses for both GWAS and IBI highlights its potential as a candidate for further explorations in PD genetics. On the other hand, rs61817862 and rs77140267, though not previously identified as PD risk loci, have shown recurring presence in the SHAP plots from both methods. This raises the possibility that these SNPs may play a role in PD's genetic framework, warranting further investigation. These findings underscore the exploratory nature of diverse analytical approaches, such as IBI, to uncover candidate genes contributing to PD, which requires validation.

The SHAP plot for the GWAS SNPs (Figure 3a) displayed several SNPs that are located in genomic regions previously associated with PD risk. Notable SNPs in this plot include rs1804972 (located at PRPF19) (Hernandez et al. 2020; Turcotte et al. 2021), rs72698636 (located at NTRK1) (Liu et al. 2018; Yang et al. 2021), and rs13097173 (located at CADPS) (Obergasteiger et al. 2017). Additionally, the GWAS SHAP plot highlights variant rs114525519, situated within the gene TRIM46. Previous studies show evidence of the involvement of TRIM family proteins in PD (Li et al. 2021) and the potential role of TRIM46 in the formation of the axon initial segment and neuronal structure (Bell et al. 2021). Furthermore, a recently published paper by Rahman et al. has proposed the following variants as potential candidates for PD: rs11772125, rs111408331, and rs72792300 (Rahman and Liu 2023). Similarly, the SHAP plot for the IBI SNPs (Figure 3b) included two more SNPs that have been previously reported to be associated with PD risk; rs117909726 (located at AXIN2) (Oosterveen et al. 2021), and rs77878665 (located at RYR3) (Klegeris et al. 2007; Sonninen et al. 2020). The IBI SHAP plot highlights several SNPs, including rs117887535, rs16887838, and rs78948272, situated at loci that have not previously been associated with PD but have known links to MS, breast cancer, and fragile bone, respectively. Some of these variants will be discussed in the next sections as compelling candidates for novel PD risk loci and are good candidates for further investigations.

Overall, the machine‐learning and SHAP analyses serve only a supporting role in this study. Their purpose is to provide complementary validation that IBI‐prioritized variants carry a stronger predictive signal and to help prioritize variants for downstream interpretation. The primary objective of this work remains the identification and prioritization of candidate PD‐associated variants or genes using IBI, relative to GWAS.

3.4. Significant Genetic Marker Candidate Discovery

In this section, we highlight candidate PD variants prioritized by IBI for further investigation. To provide context, GWAS results were used as a baseline comparison, alongside evaluation metrics, SHAP analysis, and existing knowledge from the literature. Our collected list consists of high‐ranking SNPs from IBI, which also fulfill other criteria of interest: either high GWAS ranking, prominence in SHAP analysis, potential impact on protein functionality, or proximity to genes previously implicated in PD or related neurological disorders.

Variants rs9933717 and rs77878665, located in genes linked to PD, are showcased for their substantial potential in PD research (Table 4). These variants stand out for their high rankings in both IBI and GWAS; they are also among the top 15 SNPs with the largest absolute SHAP values, indicating their significant roles in PD prediction (Table 4). The variant rs9933717 is situated near gene CDH8, which is known to be involved in pathways crucial for high‐order executive functions and cognitive deficits seen in ASDs, as well as NDDs like PD and HD (Liu et al. 2011; Friedman et al. 2015).

TABLE 4.

Key novel SNPs for PD, located within or near genes with prior PD‐related or neurologic relevance.

Variant ID MAF IBI rank GWAS rank Overlapped or nearest gene Type Knowledge on association
rs9933717* 0.02 2 113 CDH8 Intergenic ASD, HD, PD (Liu et al. 2011; Friedman et al. 2015)
rs77878665* 0.03 113 267 RYR3 Intronic NDDs, PD (Klegeris et al. 2007; Supnet et al. 2010; Gong et al. 2018; Sun and Wei 2021)
rs11986753 0.01 190 410119 ARHGEF10 3' UTR CMT, PD (Glaab and Schneider 2015; Wüllner et al. 2016; García‐Campayo et al. 2018; Bloch‐Gallego and Anderson 2023)
rs10952316 0.03 257 2137 PRKAG2 Intronic AD, PD (Bharadwaj and Martins 2020; Dulski et al. 2022; Weintraub et al. 2022)
Open in a new tab

Abbreviations: GWAS, genome‐wide association studies; IBI, Individualized Bayesian Inference; SNPs, single‐nucleotide polymorphisms.

*

indicates the SNPs are ranked within the top 15 for predicting PD by SHAP analysis.

Similarly, since the RYRs family is known for associations with NDDs, including PD, and RYR3 is associated with AD (Supnet et al. 2010; Gong et al. 2018; Sun and Wei 2021), rs77878665 could be an interesting PD candidate due to its notable high ranking by both methods and its high SHAP value ranking. Another interesting variant, rs11986753, is located in the ARHGEF10 gene that is known to contribute to neuropsychiatric disorders (schizophrenia, autism, bipolar disorder, and depression), and NDDs (CMT, AD, and PD) (Glaab and Schneider 2015; Wüllner et al. 2016; García‐Campayo et al. 2018; Bloch‐Gallego and Anderson 2023). Identified as a 3 Prime UTR variant, rs11986753 may play a notable role in regulating the ARHGEF10 expression, underscoring its potential importance in the genomic landscape of PD (Table 4). Similarly, the gene PRKAG2 that encodes a regulatory subunit of AMP‐activated protein kinase involved in cellular energy homeostasis, has been linked to AD and PD risk (Bharadwaj and Martins 2020; Dulski et al. 2022; Weintraub et al. 2022). This association makes the variant rs10952316 located in PRKAG2, a notable candidate for further investigations in PD research (Table 4).

IBI identified several high‐ranking variants located in genes not previously reported to be directly associated with PD, highlighting its ability to prioritize SNPs that warrant further investigation (Table 5). Among these, rs117887535 (within ANO2), rs61817862 (within PLEKHA6), and rs77140267 (intergenic near AVPR1A and DPY19L2) are not only ranked high by both GWAS and IBI, but they're also among the top SHAP variants (Table 5). The variant rs117887535 is located at gene ANO2, which was previously reported as an autoimmune target in MS (Ayoglu et al. 2016). Given the mutual neurodegenerative aspects between MS and PD, rs117887535 within ANO2 may be a potential risk locus for PD. Similarly, rs61817862 is situated near the gene PLEKHA6, which has been suggested to play a role in regulating copper homeostasis (Sluysmans et al. 2021). Although PLEKHA6 is not directly associated with PD, impaired copper levels have been reported to be related to various NDDs, including PD (Bisaglia and Bubacco 2020). Furthermore, the rs77140267 variant is near AVPR1A, which is involved in various neurological conditions, such as autism (Wassink et al. 2004; Israel et al. 2008) and personality traits (Walum et al. 2008; Meyer‐Lindenberg et al. 2009) related to brain function. Therefore, these loci represent biologically plausible candidates for further investigation in PD research.

TABLE 5.

Key novel SNPs for PD, located within or near genes with no clear prior evidence for PD association.

Variant ID MAF IBI rank GWAS rank Overlapped or nearest gene Type Knowledge on association
rs117887535 0.01 4 725 ANO2 Intronic MS (Ayoglu et al. 2016)
rs61817862 0.02 62 127 PLEKHA6 Intronic Copper Homeostasis (Sluysmans et al. 2021)
rs77140267 0.02 78 162 AVPR1A/DPY19L2 Intergenic ASD (Wassink et al. 2004; Israel et al. 2008; Walum et al. 2008; Meyer‐Lindenberg et al. 2009)
rs146284335 0.01 95 200371 XIRP1 Missense
rs77264713 0.03 106 11381 NKAIN3 Intronic AD (van der Plaat et al. 2018; Chai et al. 2020)
rs58880010 0.02 134 54339 FOXL2NB Missense
rs117083334 0.04 138 336100 CEP164 Missense
rs62640032 0.03 145 55079 TIMM22 Synonymous Mitochondrial function (Nicolas et al. 2019; Palmer et al. 2021)
rs111301435 0.01 146 284490 APLP2 Missense
rs61750361 0.01 191 18606 FOXL2 Synonymous
rs61744696 0.02 285 2177 KCNT1 Missense Epilepsy (Xie et al. 2021; Lu et al. 2022; Miziak and Czuczwar 2022)
Open in a new tab

Abbreviations: GWAS, genome‐wide association studies; IBI, Individualized Bayesian Inference; SNPs, single‐nucleotide polymorphisms.

Despite not being ranked highly by GWAS, several variants ranked significantly by IBI are located near genes that may share pathways with PD. These include rs77264713 [NKAIN3], rs62640032 [TIMM22], rs61750361 [FOXL2], and rs61744696 [KCNT1]. NKAIN3 is implicated in neuronal development, synaptic plasticity, and brain function, and has been associated with AD and the age of AD onset in Europeans (van der Plaat et al. 2018; Chai et al. 2020). Although limited evidence directly linking NKAIN3 to PD, the variant rs77264713, located within this gene, is notably ranked 106 by IBI, suggesting that this locus warrants further investigation. The known neurological implications of NKAIN3, including its role in brain function, make it a gene of potential interest for further PD research. Furthermore, TIMM22, which is involved in the translocation of proteins across the inner mitochondrial membrane, plays a role in mitochondrial function (Nicolas et al. 2019; Palmer et al. 2021). Mitochondrial dysfunction is a well‐established factor in the development and progression of PD, and they share several genes (such as PINK1 and Parkin) (Narendra and Youle 2011; Pilsl and Winklhofer 2012). This makes the SNP rs62640032, a synonymous variant ranked among the top 200 and located at TIMM22, an interesting candidate locus for PD.

Additionally, previous studies suggest KCNT1 as a contributing factor in epilepsy(Xie et al. 2021; Lu et al. 2022; Miziak and Czuczwar 2022). Recent research suggests that epilepsy could be a risk factor for developing PD, and there may be underlying biological connections between the two conditions (Gruntz et al. 2018). Since rs61744696 is identified as a missense variant located in the coding region of KCNT1, it'd be a compelling candidate for further investigations in PD research. This link also strengthens the case for examining KCNT1's role in PD, considering its established connection with neurological disorders. Finally, the missense variants rs146284335, rs58880010, rs117083334, and rs111301435, although not located in genes directly linked to NDDs or PD, are among the top 200 ranked by IBI, making them candidates for further investigation due to their potential functional impact on the encoded protein. In summary, this integrative and multi‐faceted analysis highlighted biologically plausible candidate loci for follow‐up studies that may improve our understanding of PD.

Our research highlights the complexity of the underlying genetic architecture of PD and emphasizes the importance of identifying potential genetic variants and loci to better understand its pathogenesis. Our analysis demonstrates the ability of IBI to prioritize SNPs with potential relevance to PD, including variants that were not highly ranked by GWAS. These findings support the use of IBI as a complementary tool for exploring the complex genetic architecture of PD. However, the highlighted loci should be regarded as candidate variants requiring further validation in independent datasets and functional studies.

Author Contributions

Jinling Liu conceived the study, designed the experiments, supervised the project, and critically revised the manuscript. Qing Lu provided important insights over the experimental design and critically edited the manuscript to help ensure the accuracy and integrity of the work. Md Asad Rahman conducted the data preprocessing and troubleshooting during various stages of the analysis. Yasaman J. Soofi conducted the GWAS and IBI analysis, ANN modeling and prediction, SHAP analysis, performed the literature search, and wrote the first draft of the manuscript. Jin Ren conducted GWAS/IBI analysis, performed prediction using ANN, SVM, and RF models, and revised the manuscript. All authors have approved the final manuscript.

Ethics Statement

This study used de‐identified data obtained through the Fox Insight Data Exploration Network (Fox DEN) under the data access policies of that resource. All participant consent and ethical oversight procedures were governed by the original Fox Insight study.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supporting File:

GEPI-50-0-s001.docx (198.1KB, docx)

Acknowledgments

This work was supported by the National Heart, Lung, and Blood Institute, National Institutes of Health [grant numbers K01HL161538, R03HL168984].

Ren, J. , Soofi Y. J., Rahman M. A., Lu Q., and Liu J.. 2026. “Individualized Bayesian Inference Identifies Novel Genetic Variants for Parkinson's Disease.” Genetic Epidemiology 50: e70045. 10.1002/gepi.70045.

Jin Ren and Yasaman J. Soofi contributed equally to this study.

Data Availability Statement

The data that support the findings of this study are openly available in The Fox Insight Data Exploration Network at https://foxden.michaeljfox.org/insight/explore/insight.jsp. The data used in this analysis came from the Fox Insight study and was obtained through the Fox Insight Data Exploration Network (Fox DEN; https://foxden.michaeljfox.org/insight/explore/insight.jsp). The following licenses and restrictions were in place: access to Fox Insight datasets is available to qualified researchers. Kindly send requests for access to these datasets to Fox DEN at https://foxden.michaeljfox.org/insight/register/genetic.

References

  1. Aarsland, D. , Creese B., Politis M., et al. 2017. “Cognitive Decline in Parkinson Disease.” Nature Reviews Neurology 13: 217–231. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Alessi, D. R. , and Sammler E.. 2018. “LRRK2 Kinase in Parkinson's Disease.” Science 360, no. 1979: 36–37. [DOI] [PubMed] [Google Scholar]
  3. Ayoglu, B. , Mitsios N., Kockum I., et al. 2016. “Anoctamin 2 Identified as an Autoimmune Target in Multiple Sclerosis.” Proceedings of the National Academy of Sciences 113: 2188–2193. [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Bell, M. , Bachmann S., Klimek J., Langerscheidt F., and Zempel H.. 2021. “Axonal TAU Sorting Requires the C‐Terminus of TAU but Is Independent of ANKG and TRIM46 Enrichment at the AIS.” Neuroscience 461: 155–171. [DOI] [PubMed] [Google Scholar]
  5. Bharadwaj, P. , and Martins R. N.. 2020. “Prkag2 Gene Expression Is Elevated and Its Protein Levels Are Associated With Increased Amyloid‐β Accumulation in the Alzheimer's Disease Brain.” Journal of Alzheimer's Disease 74: 441–448. [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Bisaglia, M. , and Bubacco L.. 2020. “Copper Ions and Parkinson's Disease: Why Is Homeostasis so Relevant?” Biomolecules 10: 195. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Bloch‐Gallego, E. , and Anderson D. I.. 2023. “Key Role of Rho GTPases in Motor Disorders Associated With Neurodevelopmental Pathologies.” Molecular Psychiatry 28: 118–126. [DOI] [PubMed] [Google Scholar]
  8. Bryant, N. , Malpeli N., Ziaee J., Blauwendraat C., Liu Z., and West A. B.. 2021. “Identification of LRRK2 Missense Variants in the Accelerating Medicines Partnership Parkinson's Disease Cohort.” Human Molecular Genetics 30: 454–466. 10.1093/hmg/ddab058. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Cai, Y. , Liu P., Xu Y., et al. 2023. “Biomarkers of Obesity‐Mediated Insulin Resistance: Focus on microRNAs.” Diabetology & Metabolic Syndrome 15: 167. [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Cereda, E. , Barichella M., Pedrolli C., et al. 2011. “Diabetes and Risk of Parkinson's Disease.” Diabetes Care 34: 2614–2623. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Chai, J.‐F. , Raichur S., Khor I. W., et al. 2020. “Associations With Metabolites in Chinese Suggest New Metabolic Roles in Alzheimer's and Parkinson's Diseases.” Human Molecular Genetics 29: 189–201. [DOI] [PubMed] [Google Scholar]
  12. Chang, D. , Nalls M. A., Hallgrímsdóttir I. B., et al. 2017. “A Meta‐Analysis of Genome‐Wide Association Studies Identifies 17 New Parkinson's Disease Risk Loci.” Nature Genetics 49: 1511–1516. 10.1038/ng.3955. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Chaudhuri, K. R. , and Schapira A. H.. 2009. “Non‐Motor Symptoms of Parkinson's Disease: Dopaminergic Pathophysiology and Treatment.” Lancet Neurology 8: 464–474. [DOI] [PubMed] [Google Scholar]
  14. Dauer, W. , and Przedborski S.. 2003. “Parkinson's Disease.” Neuron 39: 889–909. [DOI] [PubMed] [Google Scholar]
  15. Dorsey, E. R. , Elbaz A., Nichols E., et al. 2018. “Global, Regional, and National Burden of Parkinson's Disease, 1990–2016: A Systematic Analysis for the Global Burden of Disease Study 2016.” Lancet Neurology 17: 939–953. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Doty, R. L. , Shaman P., Applebaum S. L., Giberson R., Siksorski L., and Rosenberg L.. 1984. “Smell Identification Ability: Changes With Age.” Science 226, no. 1979: 1441–1443. [DOI] [PubMed] [Google Scholar]
  17. Dulski, J. , Uitti R. J., Ross O. A., and Wszolek Z. K.. 2022. “Genetic Architecture of Parkinson's Disease Subtypes–Review of the Literature.” Frontiers in Aging Neuroscience 14: 1023574. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Eura, N. , Ogasawara M., and Nishino I.. 2024. “Recent Topics of Oculopharyngodistal Myopathy.” Neurology and Clinical Neuroscience 12: 10–15. [Google Scholar]
  19. Fernández‐Santiago, R. , Martín‐Flores N., Antonelli F., et al. 2019. “SNCA and mTOR Pathway Single Nucleotide Polymorphisms Interact to Modulate the Age at Onset of Parkinson's Disease.” Movement Disorders 34: 1333–1344. [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Friedman, L. G. , Riemslagh F. W., Sullivan J. M., et al. 2015. “Cadherin‐8 Expression, Synaptic Localization, and Molecular Control of Neuronal Form in Prefrontal Corticostriatal Circuits.” Journal of Comparative Neurology 523: 75–92. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Gagnon, J.‐F. , Postuma R. B., Mazza S., Doyon J., and Montplaisir J.. 2006. “Rapid‐Eye‐Movement Sleep Behaviour Disorder and Neurodegenerative Diseases.” Lancet Neurology 5: 424–432. [DOI] [PubMed] [Google Scholar]
  22. García‐Campayo, J. , Puebla‐Guedea M., Labarga A., et al. 2018. “Epigenetic Response to Mindfulness in Peripheral Blood Leukocytes Involves Genes Linked to Common Human Diseases.” Mindfulness 9: 1146–1159. [Google Scholar]
  23. Gilks, W. P. , Abou‐Sleiman P. M., Gandhi S., et al. 2005. “A Common LRRK2 Mutation in Idiopathic Parkinson's Disease.” Lancet 365: 415–416. [DOI] [PubMed] [Google Scholar]
  24. Glaab, E. , and Schneider R.. 2015. “Comparative Pathway and Network Analysis of Brain Transcriptome Changes During Adult Aging and in Parkinson's Disease.” Neurobiology of Disease 74: 1–13. [DOI] [PubMed] [Google Scholar]
  25. Goldstein, D. S. 2006. “Orthostatic Hypotension as an Early Finding in Parkinson's Disease.” Clinical Autonomic Research 16: 46–54. [DOI] [PubMed] [Google Scholar]
  26. Gong, S. , Su B. B., Tovar H., et al. 2018. “Polymorphisms Within RYR3 Gene Are Associated With Risk and Age at Onset of Hypertension, Diabetes, and Alzheimer's Disease.” American Journal of Hypertension 31: 818–826. [DOI] [PubMed] [Google Scholar]
  27. Greenbaum, L. , and Lerer B.. 2015. “Pharmacogenetics of Antipsychotic‐Induced Movement Disorders as a Resource for Better Understanding Parkinson's Disease Modifier Genes.” Frontiers in Neurology 6: 27. [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Greenbaum, L. , Smith R. C., Lorberboym M., et al. 2012. “Association of the ZFPM2 Gene With Antipsychotic‐Induced Parkinsonism in Schizophrenia Patients.” Psychopharmacology 220: 519–528. [DOI] [PubMed] [Google Scholar]
  29. Gruntz, K. , Bloechliger M., Becker C., et al. 2018. “Parkinson Disease and the Risk of Epileptic Seizures.” Annals of Neurology 83: 363–374. [DOI] [PubMed] [Google Scholar]
  30. Henden, L. , Fearnley L. G., and Southwood D., et al. 2024. “Short Tandem Repeat Expansions in LRP12 Are Absent in Cohorts of Familial and Sporadic Amyotrophic Lateral Sclerosis Patients of European Ancestry.” Amyotroph Lateral Scler Frontotemporal Degener 25: 644–647. [DOI] [PubMed] [Google Scholar]
  31. Hendrickx, D. M. , and Glaab E.. 2020. “Comparative Transcriptome Analysis of Parkinson's Disease and Hutchinson‐Gilford Progeria Syndrome Reveals Shared Susceptible Cellular Network Processes.” BMC Medical Genomics 13: 114. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Hernandez, S. M. , Tikhonova E. B., and Karamyshev A. L.. 2020. “Protein‐Protein Interactions in Alpha‐Synuclein Biogenesis: New Potential Targets in Parkinson's Disease.” Frontiers in Aging Neuroscience 12: 72. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Huang, S. , Zhang C., Xie X., et al. 2022. “GRID2 Aberration Leads to Disturbance in Neuroactive Ligand‐Receptor Interactions via Changes to the Species Richness and Composition of Gut Microbes.” Biochemical and Biophysical Research Communications 631: 9–17. [DOI] [PubMed] [Google Scholar]
  34. Infante, J. , Prieto C., Sierra M., et al. 2015. “Identification of Candidate Genes for Parkinson's Disease Through Blood Transcriptome Analysis in LRRK2‐G2019S Carriers, Idiopathic Cases, and Controls.” Neurobiology of Aging 36: 1105–1109. [DOI] [PubMed] [Google Scholar]
  35. Israel, S. , Lerer E., Shalev I., et al. 2008. “Molecular Genetic Studies of the Arginine Vasopressin 1a Receptor (AVPR1a) and the Oxytocin Receptor (OXTR) in Human Behaviour: From Autism to Altruism With Some Notes in Between.” Progress in Brain Research 170: 435–449. [DOI] [PubMed] [Google Scholar]
  36. Jankovic, J. 2008. “Parkinson's Disease: Clinical Features and Diagnosis.” Journal of Neurology, Neurosurgery & Psychiatry 79: 368–376. [DOI] [PubMed] [Google Scholar]
  37. Jankovic, J. , Schwartz K. S., and Ondo W.. 1999. “Re‐Emergent Tremor of Parkinson's Disease.” Journal of Neurology, Neurosurgery & Psychiatry 67: 646–650. [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Klegeris, A. , Choi H. B., McLarnon J. G., and McGeer P. L.. 2007. “Functional Ryanodine Receptors Are Expressed by Human Microglia and THP‐1 Cells: Their Possible Involvement in Modulation of Neurotoxicity.” Journal of Neuroscience Research 85: 2207–2215. [DOI] [PubMed] [Google Scholar]
  39. Korte, A. , and Farlow A.. 2013. “The Advantages and Limitations of Trait Analysis With GWAS: A Review.” Plant Methods 9: 29. [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Kume, K. , Kurashige T., Muguruma K., et al. 2023. “CGG Repeat Expansion in LRP12 in Amyotrophic Lateral Sclerosis.” American Journal of Human Genetics 110: 1086–1097. [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. Lai, D. , Alipanahi B., Fontanillas P., et al. 2021. “Genomewide Association Studies of LRRK2 Modifiers of Parkinson's Disease.” Annals of Neurology 90: 76–88. [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Lesage, S. , Patin E., Condroyer C., et al. 2010. “Parkinson's Disease‐Related LRRK2 G2019S Mutation Results From Independent Mutational Events in Humans.” Human Molecular Genetics 19: 1998–2004. 10.1093/hmg/ddq081. [DOI] [PubMed] [Google Scholar]
  43. Li, C. , Ou R., Hou Y., et al. 2021. “Genetic Analysis of TRIM Family Genes for Early‐Onset Parkinson's Disease in Chinese Population.” Parkinsonism & Related Disorders 90: 105–113. [DOI] [PubMed] [Google Scholar]
  44. Liu, K. , Zhang W., Li Y., et al. 2018. “Human Peripheral Blood‐Derived Mesenchymal Stem Cells With NTRK1 Over‐Expression Enhance Repairing Capability in a Rat Model of Parkinson's Disease.” Cytotechnology 70: 1291–1299. [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. Liu, S. , Zhang Z., Li L., et al. 2021. “ADAM10‐and Presenilin 1/γ‐secretase‐dependent Cleavage of PTPRT Mitigates Neurodegeneration of Alzheimer's Disease.” bioRxiv: 2021.01.11.426157. [Google Scholar]
  46. Liu, X. , Cheng R., Verbitsky M., et al. 2011. “Genome‐Wide Association Study Identifies Candidate Genes for Parkinson's Disease in an Ashkenazi Jewish Population.” BMC Medical Genetics 12: 104. [DOI] [PMC free article] [PubMed] [Google Scholar]
  47. Lu, J. , Zhao G., Lv D., Cao L., and Zhao G.. 2022. “Autosomal Dominant Sleep‐Related Hypermotor Epilepsy Associated With a Novel Mutation of KCNT1.” Translational Neuroscience 13: 240–245. [DOI] [PMC free article] [PubMed] [Google Scholar]
  48. Lu, Y. , Chen W., Wei C., Zhu Y., and Xu R.. 2021. “Potential Common Genetic Risks of Sporadic Parkinson's Disease and Amyotrophic Lateral Sclerosis in the Han Population of Mainland China.” Frontiers in Neuroscience 15: 753870. [DOI] [PMC free article] [PubMed] [Google Scholar]
  49. Lubbe, S. J. , Bustos B. I., Hu J., et al. 2021. “Assessing the Relationship Between Monoallelic PRKN Mutations and Parkinson's Risk.” Human Molecular Genetics 30: 78–86. 10.1093/hmg/ddaa273. [DOI] [PMC free article] [PubMed] [Google Scholar]
  50. Marras, C. , Beck J. C., Bower J. H., et al. 2018. “Prevalence of Parkinson's Disease Across North America.” npj Parkinson's Disease 4: 21. [DOI] [PMC free article] [PubMed] [Google Scholar]
  51. Martin, I. , Kim J. W., Dawson V. L., and Dawson T. M.. 2014. “LRRK2 Pathobiology in Parkinson's Disease.” Journal of Neurochemistry 131: 554–565. [DOI] [PMC free article] [PubMed] [Google Scholar]
  52. Martín‐Flores, N. , Fernández‐Santiago R., Antonelli F., et al. 2019. “MTOR Pathway‐Based Discovery of Genetic Susceptibility to L‐DOPA‐Induced Dyskinesia in Parkinson's Disease Patients.” Molecular Neurobiology 56: 2092–2100. [DOI] [PubMed] [Google Scholar]
  53. Mata, I. F. 2004. “Parkin Genetics: One Model for Parkinson's Disease.” Human Molecular Genetics 13: 127R–133R. [DOI] [PubMed] [Google Scholar]
  54. Mata, I. F. , Wedemeyer W. J., Farrer M. J., Taylor J. P., and Gallo K. A.. 2006. “LRRK2 in Parkinson's Disease: Protein Domains and Functional Insights.” Trends in Neurosciences 29: 286–293. [DOI] [PubMed] [Google Scholar]
  55. Meyer‐Lindenberg, A. , Kolachana B., Gold B., et al. 2009. “Genetic Variants in AVPR1A Linked to Autism Predict Amygdala Activation and Personality Traits in Healthy Humans.” Molecular Psychiatry 14: 968–975. [DOI] [PMC free article] [PubMed] [Google Scholar]
  56. Miziak, B. , and Czuczwar S. J.. 2022. “Approaches for the Discovery of Drugs That Target K Na 1.1 Channels in KCNT1‐Associated Epilepsy.” Expert Opinion on Drug Discovery 17: 1313–1328. [DOI] [PubMed] [Google Scholar]
  57. Mortezaei, Z. , Lanjanian H., and Masoudi‐Nejad A.. 2017. “Candidate Novel Long Noncoding RNAs, Micrornas and Putative Drugs for Parkinson's Disease Using a Robust and Efficient Genome‐Wide Association Study.” Genomics 109: 158–164. [DOI] [PubMed] [Google Scholar]
  58. Murugesan, V. , and Rajagopal S.. 2022. “Regulatory Role of ADGRL3, PARK2, and CNTNAP2 in Neurodevelopmental Disorders.” In Proteins Associated With Neurodevelopmental Disorders, 291–299. Springer. [Google Scholar]
  59. Nalls, M. A. , Blauwendraat C., Vallerga C. L., et al. 2019. “Identification of Novel Risk Loci, Causal Insights, and Heritable Risk for Parkinson's Disease: A Meta‐Analysis of Genome‐Wide Association Studies.” Lancet Neurology 18: 1091–1102. [DOI] [PMC free article] [PubMed] [Google Scholar]
  60. Nalls, M. A. , Pankratz N., Lill C. M., et al. 2014. “Large‐Scale Meta‐Analysis of Genome‐Wide Association Data Identifies Six New Risk Loci for Parkinson's Disease.” Nature Genetics 46: 989–993. [DOI] [PMC free article] [PubMed] [Google Scholar]
  61. Narendra, D. P. , and Youle R. J.. 2011. “Targeting Mitochondrial Dysfunction: Role for PINK1 and Parkin in Mitochondrial Quality Control.” Antioxidants & Redox Signaling 14: 1929–1938. [DOI] [PMC free article] [PubMed] [Google Scholar]
  62. Nicolas, E. , Tricarico R., Savage M., Golemis E. A., and Hall M. J.. 2019. “Disease‐Associated Genetic Variation in Human Mitochondrial Protein Import.” American Journal of Human Genetics 104: 784–801. [DOI] [PMC free article] [PubMed] [Google Scholar]
  63. Nutt, J. G. , and Wooten G. F.. 2005. “Diagnosis and Initial Management of Parkinson's Disease.” New England Journal of Medicine 353: 1021–1027. [DOI] [PubMed] [Google Scholar]
  64. Nuytemans, K. , Theuns J., Cruts M., and Van Broeckhoven C.. 2010. “Genetic Etiology of Parkinson Disease Associated With Mutations in the SNCA, PARK2, PINK1, PARK7, and LRRK2 Genes: A Mutation Update.” Human Mutation 31: 763–780. [DOI] [PMC free article] [PubMed] [Google Scholar]
  65. Obergasteiger, J. , Überbacher C., Pramstaller P. P., Hicks A. A., Corti C., and Volta M.. 2017. “CADPS2 Gene Expression Is Oppositely Regulated by LRRK2 and Alpha‐Synuclein.” Biochemical and Biophysical Research Communications 490: 876–881. [DOI] [PubMed] [Google Scholar]
  66. Oosterveen, T. , Garção P., Moles‐Garcia E., et al. 2021. “Pluripotent Stem Cell Derived Dopaminergic Subpopulations Model the Selective Neuron Degeneration in Parkinson's Disease.” Stem Cell Reports 16: 2718–2735. [DOI] [PMC free article] [PubMed] [Google Scholar]
  67. Palmer, C. S. , Anderson A. J., and Stojanovski D.. 2021. “Mitochondrial Protein Import Dysfunction: Mitochondrial Disease, Neurodegenerative Disease and Cancer.” FEBS Letters 595: 1107–1131. [DOI] [PubMed] [Google Scholar]
  68. Pan, C. , Li M., Wang J., et al. 2022. “miR‐4431 Targets TRIP10/PRKD1 and Impairs Glucose Metabolism.” Journal of Diabetes Investigation 13: 617–627. [DOI] [PMC free article] [PubMed] [Google Scholar]
  69. Parkinson's Foundation . 2023. https://www.parkinson.org/understanding-parkinsons/statistics. https://www.parkinson.org/understanding-parkinsons/statistics. Accessed 14 Feb 2023.
  70. Pilsl, A. , and Winklhofer K. F.. 2012. “Parkin, PINK1 and Mitochondrial Integrity: Emerging Concepts of Mitochondrial Dysfunction in Parkinson's Disease.” Acta Neuropathologica 123: 173–188. [DOI] [PubMed] [Google Scholar]
  71. van der Plaat, D. A. , de Jong K., de Vries M., et al. 2018. “Occupational Exposure to Pesticides Is Associated With Differential DNA Methylation.” Occupational and Environmental Medicine 75: 427–435. [DOI] [PMC free article] [PubMed] [Google Scholar]
  72. Polymeropoulos, M. H. , Lavedan C., Leroy E., et al. 1997. “Mutation in the α‐synuclein Gene Identified in Families With Parkinson's Disease.” Science 276, no. 1979: 2045–2047. [DOI] [PubMed] [Google Scholar]
  73. Rahman, M. A. , Cai C., Bo N., et al. 2023. “An Individualized Bayesian Method for Estimating Genomic Variants of Hypertension.” BMC Genomics 23: 863. [DOI] [PMC free article] [PubMed] [Google Scholar]
  74. Rahman, M. A. , and Liu J.. 2023. “A Genome‐Wide Association Study Coupled With Machine Learning Approaches to Identify Influential Demographic and Genomic Factors Underlying Parkinson's Disease.” Frontiers in Genetics 14: 1230579. [DOI] [PMC free article] [PubMed] [Google Scholar]
  75. Sanchez‐Contreras, M. Y. , Kouri N., Cook C. N., et al. 2018. “Replication of Progressive Supranuclear Palsy Genome‐Wide Association Study Identifies SLCO1A2 and DUSP10 as New Susceptibility Loci.” Molecular Neurodegeneration 13: 37. [DOI] [PMC free article] [PubMed] [Google Scholar]
  76. Sandyk, R. 1993. “The Relationship Between Diabetes Mellitus and Parkinson's Disease.” International Journal of Neuroscience 69: 125–130. [DOI] [PubMed] [Google Scholar]
  77. Satake, W. , Nakabayashi Y., Mizuta I., et al. 2009. “Genome‐Wide Association Study Identifies Common Variants at Four Loci as Genetic Risk Factors for Parkinson's Disease.” Nature Genetics 41: 1303–1307. [DOI] [PubMed] [Google Scholar]
  78. Seo, S. H. , Bacolla A., Yoo D., et al. 2020. “Replication‐Based Rearrangements Are a Common Mechanism for SNCA Duplication in Parkinson's Disease.” Movement Disorders 35: 868–876. [DOI] [PMC free article] [PubMed] [Google Scholar]
  79. Shimizu, T. , Ishiura H., Hara M., et al. 2022. “Expanded Clinical Spectrum of Oculopharyngodistal Myopathy Type 1.” Muscle & Nerve 66: 679–685. [DOI] [PubMed] [Google Scholar]
  80. Singleton, A. , and Hardy J.. 2019. “Progress in the Genetic Analysis of Parkinson's Disease.” Human Molecular Genetics 28: R215–R218. [DOI] [PMC free article] [PubMed] [Google Scholar]
  81. Singleton, A. B. , Farrer M. J., and Bonifati V.. 2013. “The Genetics of P Arkinson's Disease: Progress and Therapeutic Implications.” Movement Disorders 28: 14–23. [DOI] [PMC free article] [PubMed] [Google Scholar]
  82. Sluysmans, S. , Méan I., Xiao T., et al. 2021. “PLEKHA5, PLEKHA6, and PLEKHA7 Bind to PDZD11 to Target the Menkes ATPase ATP7A to the Cell Periphery and Regulate Copper Homeostasis.” Molecular Biology of the Cell 32: ar34. [DOI] [PMC free article] [PubMed] [Google Scholar]
  83. Sonninen, T.‐M. , Hämäläinen R. H., Koskuvi M., et al. 2020. “Metabolic Alterations in Parkinson's Disease Astrocytes.” Scientific Reports 10: 14474. [DOI] [PMC free article] [PubMed] [Google Scholar]
  84. La Spada, A. , and Ranum L. P. W.. 2010. “Molecular Genetic Advances in Neurological Disease: Special Review Issue.” Human Molecular Genetics 19: R1–R3. [DOI] [PMC free article] [PubMed] [Google Scholar]
  85. Sun, L. , and Wei H.. 2021. “Ryanodine Receptors: A Potential Treatment Target in Various Neurodegenerative Disease.” Cellular and Molecular Neurobiology 41: 1613–1624. [DOI] [PMC free article] [PubMed] [Google Scholar]
  86. Supnet, C. , Noonan C., Richard K., Bradley J., and Mayne M.. 2010. “Up‐Regulation of the Type 3 Ryanodine Receptor Is Neuroprotective in the TgCRND8 Mouse Model of Alzheimer's Disease.” Journal of Neurochemistry 112: 356–365. [DOI] [PubMed] [Google Scholar]
  87. Turcotte, M.‐A. , Garant J.‐M., Cossette‐Roberge H., and Perreault J.‐P.. 2021. “Guanine Nucleotide‐Binding Protein‐Like 1 (GNL1) Binds RNA G‐Quadruplex Structures in Genes Associated With Parkinson's Disease.” RNA Biology 18: 1339–1353. [DOI] [PMC free article] [PubMed] [Google Scholar]
  88. Usenko, T. , Bezrukova A., Basharova K., et al. 2021. “Comparative Transcriptome Analysis in Monocyte‐Derived Macrophages of Asymptomatic GBA Mutation Carriers and Patients With GBA‐Associated Parkinson's Disease.” Genes 12: 1545. [DOI] [PMC free article] [PubMed] [Google Scholar]
  89. Valente, E. M. , Abou‐Sleiman P. M., Caputo V., et al. 2004. “Hereditary Early‐Onset Parkinson's Disease Caused by Mutations in PINK1.” Science 304, no. 1979: 1158–1160. [DOI] [PubMed] [Google Scholar]
  90. Walum, H. , Westberg L., Henningsson S., et al. 2008. “Genetic Variation in the Vasopressin Receptor 1a Gene (AVPR1A) Associates With Pair‐Bonding Behavior in Humans.” Proceedings of the National Academy of Sciences 105: 14153–14156. [DOI] [PMC free article] [PubMed] [Google Scholar]
  91. Wassink, T. H. , Piven J., Vieland V. J., et al. 2004. “Examination of AVPR1a as An Autism Susceptibility Gene.” Molecular Psychiatry 9: 968–972. [DOI] [PubMed] [Google Scholar]
  92. Weintraub, D. , Posavi M., Fontanillas P., et al. 2022. “Genetic Prediction of Impulse Control Disorders in Parkinson's Disease.” Annals of Clinical and Translational Neurology 9: 936–949. [DOI] [PMC free article] [PubMed] [Google Scholar]
  93. Wilczyński, J. , Pedrycz A., Mucha D., Ambroży T., and Mucha D.. 2017. “Body Posture, Postural Stability, and Metabolic Age in Patients With Parkinson's Disease.” BioMed Research International 2017: 1–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  94. Wüllner, U. , Kaut O., deBoni L., Piston D., and Schmitt I.. 2016. “DNA Methylation in Parkinson's Disease.” Journal of Neurochemistry 139: 108–120. [DOI] [PubMed] [Google Scholar]
  95. Xie, N. , Qin W., Deng J., et al. 2021. “A Novel KCNT1 Mutation in a Chinese Family With Severe Autosomal‐Dominant Nocturnal Frontal Lobe Epilepsy.” Translational Neuroscience 12: 330–334. [DOI] [PMC free article] [PubMed] [Google Scholar]
  96. Xu, Q. , Park Y., Huang X., et al. 2011. “Diabetes and Risk of Parkinson's Disease.” Diabetes Care 34: 910–915. [DOI] [PMC free article] [PubMed] [Google Scholar]
  97. Yang, W. , Hao W., Meng Z., et al. 2021. “Molecular Regulatory Mechanism and Toxicology of Neurodegenerative Processes in MPTP/Probenecid‐Induced Progressive Parkinson's Disease Mice Model Revealed by Transcriptome.” Molecular Neurobiology 58: 603–616. [DOI] [PMC free article] [PubMed] [Google Scholar]
  98. Zheng, R. , Qiao S., Chen Y., et al. 2021. “Association Analysis and Polygenic Risk Score Evaluation of 38 GWAS‐Identified Loci in a Chinese Population With Parkinson's Disease.” Neuroscience Letters 762: 136150. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supporting File:

GEPI-50-0-s001.docx (198.1KB, docx)

Data Availability Statement

The data that support the findings of this study are openly available in The Fox Insight Data Exploration Network at https://foxden.michaeljfox.org/insight/explore/insight.jsp. The data used in this analysis came from the Fox Insight study and was obtained through the Fox Insight Data Exploration Network (Fox DEN; https://foxden.michaeljfox.org/insight/explore/insight.jsp). The following licenses and restrictions were in place: access to Fox Insight datasets is available to qualified researchers. Kindly send requests for access to these datasets to Fox DEN at https://foxden.michaeljfox.org/insight/register/genetic.

Articles from Genetic Epidemiology are provided here courtesy of Wiley

RESOURCES