Structural characterization of rare missense variants within known neurodegenerative disease proteins: Genetics/omics and systems biology. (7th December 2020)

• Record Type: Journal Article
• Authors: Palmer, Ellen L.; Moth, Christopher; Benchek, Penelope; Wheeler, Nicholas R.; Kunkle, Brian W.; Hamilton‐Nelson, Kara L.; Griswold, Anthony J.; Naj, Adam C.; Farrer, Lindsay A.; Martin, Eden R.; Pericak‐Vance, Margaret A.; Haines, Jonathan L.; Sheehan, Jonathan H.; Capra, John A.; Bush, William S.
• Is Part Of: Alzheimer's & dementia. Volume 16(2020)Supplement 2
• Page Start: n/a
• Page End: n/a
• Publication Date: 2020-12-07
• Subjects: Alzheimer's disease -- Periodicals; Alzheimer Disease -- Periodicals; Dementia -- Periodicals; Démence; Maladie d'Alzheimer; Périodique électronique (Descripteur de forme); Ressource Internet (Descripteur de forme); 616.83
• DOI: 10.1002/alz.046405
• Languages: English
• ISSNs: 1552-5260

Abstract

Abstract: Background: Recent work suggests unexplained heritability of Alzheimer's Disease (AD) may involve rare variants in genes implicated in other neurodegenerative disorders. However, existing gene‐based tests have insufficient power to detect true associations when neutral variants are included. Therefore, identifying variants with potentially high impact on protein function before computing gene‐based statistical associations is critical. Method: We employed the Phenotype Consensus ANalysis (PCAN) approach to identify genes with Mendelian associations to neurodegenerative disorders using semantic similarity, resulting in a candidate gene list of 24 genes not associated with AD and two previously associated genes ( PSEN1 and TREM2 ). We applied the PathProx algorithm to identify Alzheimer's Disease Sequencing Project (ADSP) Discovery Phase variants significantly clustered with known pathogenic variants compared to benign variants within 3D protein structures. Pathogenic variants for Mendelian neurodegenerative disorders were curated from ClinVar and benign variants from ExAC. Additionally, we predicted the impact of each ADSP variant on protein stability (vs. canonical wildtype amino acid) via change in free energy of folding (∆∆G) estimated by Rosetta software. Result: Out of 1, 365 ADSP missense variants in 26 genes, we identified 230 that were proximal (using PathProx) to known pathogenic variants. A total of 230 variants had ∆∆G values indicating their potential to thermodynamically destabilize a protein's structure (91 variants were also proximal to known clinical variants). Gene‐based testing using these predicted functional variants identified two novel genes ( CSF1R, and SERPINI1 ) associated with AD risk. Finally, models for PSEN1 and TREM2 using the PathProx or ∆∆G‐identified variants remained statistically significant with similar p‐values to models using all missense variants. Conclusion: We have identified subsets of missense variants likely to impact the function of proteins associated with neurodegenerative diseases in the ADSP. These annotations can be used in multiple ways, such as highlighting variant sets for use in gene‐based testing and prioritizing variants for in vitro testing. Additionally, functional characterization of rare variants using PathProx and ∆∆G could aide in in heterogeneity testing of multifactorial diseases like AD. Finally, our methodology is phenotype‐centric and applicable to other diseases with overlapping phenotypes to identify candidate genes in a variety of neurological disorders.