A telescope GWAS analysis strategy, based on SNPs-genes-pathways ensamble and on multivariate algorithms, to characterize late onset Alzheimer’s disease

Margherita Squillario, Giulia Abate, Federico Tomasi, Veronica Tozzo, Annalisa Barla, Daniela Uberti, Michael W. Weiner, Paul Aisen, Ronald Petersen, Jack Clifford, William J. Jagust, John Q. Trojanowki, Arthur W. Toga, Laurel Beckett, Robert C. Green, Andrew J. Saykin, John C. Morris, Pilhan Kim, Zaven S. Khachaturian, Greg Sorensen, Maria Carrillo, Lew Kuller, Marc Raichle, Steven M. Paul, Peter Davies, Howard Fillit, Franz Hefti, Davie Holtzman, Marek M. Mesulam, William Z. Potter, Peter J. Snyder, Tom Montine, Ronald G. Thomas, Michael Donohue, Sarah Walter, Tamie Sather, Gus Jiminez, Archana B. Balasubramanian, Jennifer Mason, Iris Sim, Danielle Harvey, Matt A. Bernstein, Nick C. Fox, Paul M. Thompson, Norbert Schuff, Charles DeCarli, Bret Borowski, Jeff Gunter, Matthew L. Senjem, Prashanthi Vemuri, David T. Jones, Kejal Kantarci, Chad Ward, Robert A. Koeppe, Norm Foster, Eric M. Reiman, Kewei Chen, Chester A. Mathis, Susan Landau, Nigel J. Cairns, Erin Householder, Lisa Taylor‐Reinwald, Virginia Lee, Magdalena Korecka, Michal Figurski, Karen Crawford, Scott Neu, Tatiana Foroud, Steven Potkin, Li Shen, Kelley Faber, Sungeun Kim, Lean Tha, Richard Frank, John Hsiao, Jeffrey Kaye, Joseph Quinn, Lisa C. Silbert, Betty Lind, Raina Carter, Sara Dolen, Beau M. Ances, Maria Carroll, Mary L. Creech, Erin Franklin, Mark A. Mintun, Stacy Schneider, Angela Oliver, Lon S. Schneider, Sonia Pawluczyk, Mauricio Beccera, Liberty Teodoro, Bryan M. Spann, James B. Brewer, Helen Vanderswag, Adam Fleisher, Daniel Marson, Randall Griffith, David W. Clark, David GeldmacherScientific 2020 · DOI

Abstract

Genome-wide association studies (GWAS) have revealed many susceptibility genes for Alzheimer's disease (AD), yet only the APOE gene has been unequivocally validated. This study proposes a "telescope" GWAS analysis strategy using multivariate machine learning to identify correlated variables across three biological levels: SNPs, genes, and pathways. By analyzing the ADNI-1 and ADNI-2 datasets, the researchers identified robust genetic signatures, confirmed the role of TOMM40, and proposed GRM7 as a novel candidate gene for AD.

Introduction: The "Missing Heritability" of LOAD

Late-Onset Alzheimer’s Disease (LOAD) is a complex neurodegenerative disorder with an estimated heritability of 60–80%. Despite the success of traditional GWAS, much of this heritability remains "missing." Standard univariate methods test one Single Nucleotide Polymorphism (SNP) at a time, often failing to capture the interconnected networks of low-penetrance variants that likely drive the disease. To address this, a multivariate approach is required to identify correlations across different biological scales.

The Telescope Methodology

The researchers developed a "telescope" approach to analyze genetic data at three interconnected levels:

  1. SNP-based: Utilizing the PALLADIO framework ($l_1l_2FS$) to select discriminative markers while filtering noise.
  2. Gene-based: Employing the Sequence Kernel Association Test (SKAT) to evaluate the aggregate effect of variants within a gene.
  3. Pathway-based: Using Group Lasso with overlap to identify relevant biological pathways, accounting for genes that participate in multiple functional networks.
Workflow of the telescope GWAS analysis strategy across three biological levels.

Experimental Setup and Classification Tasks

The study utilized two independent, unimputed datasets from the Alzheimer’s Disease Neuroimaging Initiative (ADNI-1 and ADNI-2). Two primary classification tasks were addressed:

  • Cases vs. Controls: Distinguishing AD patients from healthy individuals.
  • APOEε4 Task: Classifying individuals based on their APOEε4 allelic status (high vs. low risk). This "filter" helps uncover polygenic profiles that might otherwise be overshadowed by the dominant effect of the APOE gene.

Key Findings: Validation and Discovery

The telescope approach successfully validated the involvement of the TOMM40 gene, specifically within the mitochondrial protein import pathway. More significantly, GRM7 (encoding metabotropic glutamate receptor 7) emerged as a novel candidate gene associated with AD across both datasets and multiple levels of analysis.

Performance metrics and SNP signatures for ADNI-1 tasks, showing balanced accuracy and correlation coefficients.

In terms of performance, the multivariate models achieved high predictive accuracy. For instance, in the ADNI-2 APOEε4 task, the SNP-based analysis of chromosome 19 reached a Balanced Accuracy of 0.94. The study also highlighted key pathways such as "GPCR ligand binding" and "detoxification of reactive oxygen species" as consistently relevant to AD risk.

Conclusion and Future Directions

The "telescope" strategy demonstrates the power of multivariate machine learning in uncovering the complex genetic architecture of LOAD. While the current study avoided imputation to test cross-platform robustness, future work will involve analyzing larger, imputed datasets to further refine these genetic signatures and integrate clinical covariates for improved predictive power.

Summary updated

Abstract

Genome-wide association studies (GWAS) have revealed a plethora of putative susceptibility genes for Alzheimer's disease (AD), with the sole exception of APOE gene unequivocally validated in independent study. Considering that the etiology of complex diseases like AD could depend on functional multiple genes interaction network, here we proposed an alternative GWAS analysis strategy based on (i) multivariate methods and on a (ii) telescope approach, in order to guarantee the identification of correlated variables, and reveal their connections at three biological connected levels. Specifically as multivariate methods, we employed two machine learning algorithms and a genetic association test and we considered SNPs, Genes and Pathways features in the analysis of two public GWAS dataset (ADNI-1 and ADNI-2). For each dataset and for each feature we addressed two binary classifications tasks: cases vs. controls and the low vs. high risk of developing AD considering the allelic status of APOEe4. This complex strategy allowed the identification of SNPs, genes and pathways lists statistically robust and meaningful from the biological viewpoint. Among the results, we confirm the involvement of TOMM40 gene in AD and we propose GRM7 as a novel gene significantly associated with AD.
BibTeX
@inproceedings{squillario2020telescope,
  author = {Margherita Squillario and Giulia Abate and Federico Tomasi and Veronica Tozzo and Annalisa Barla and Daniela Uberti and Michael W. Weiner and Paul Aisen and Ronald Petersen and Jack Clifford and William J. Jagust and John Q. Trojanowki and Arthur W. Toga and Laurel Beckett and Robert C. Green and Andrew J. Saykin and John C. Morris and Pilhan Kim and Zaven S. Khachaturian and Greg Sorensen and Maria Carrillo and Lew Kuller and Marc Raichle and Steven M. Paul and Peter Davies and Howard Fillit and Franz Hefti and Davie Holtzman and Marek M. Mesulam and William Z. Potter and Peter J. Snyder and Tom Montine and Ronald G. Thomas and Michael Donohue and Sarah Walter and Tamie Sather and Gus Jiminez and Archana B. Balasubramanian and Jennifer Mason and Iris Sim and Danielle Harvey and Matt A. Bernstein and Nick C. Fox and Paul M. Thompson and Norbert Schuff and Charles DeCarli and Bret Borowski and Jeff Gunter and Matthew L. Senjem and Prashanthi Vemuri and David T. Jones and Kejal Kantarci and Chad Ward and Robert A. Koeppe and Norm Foster and Eric M. Reiman and Kewei Chen and Chester A. Mathis and Susan Landau and Nigel J. Cairns and Erin Householder and Lisa Taylor‐Reinwald and Virginia Lee and Magdalena Korecka and Michal Figurski and Karen Crawford and Scott Neu and Tatiana Foroud and Steven Potkin and Li Shen and Kelley Faber and Sungeun Kim and Lean Tha and Richard Frank and John Hsiao and Jeffrey Kaye and Joseph Quinn and Lisa C. Silbert and Betty Lind and Raina Carter and Sara Dolen and Beau M. Ances and Maria Carroll and Mary L. Creech and Erin Franklin and Mark A. Mintun and Stacy Schneider and Angela Oliver and Lon S. Schneider and Sonia Pawluczyk and Mauricio Beccera and Liberty Teodoro and Bryan M. Spann and James B. Brewer and Helen Vanderswag and Adam Fleisher and Daniel Marson and Randall Griffith and David W. Clark and David Geldmacher},
  title = {{A telescope GWAS analysis strategy, based on SNPs-genes-pathways ensamble and on multivariate algorithms, to characterize late onset Alzheimer’s disease}},
  booktitle = {Scientific},
  year = {2020},
  doi = {10.1038/s41598-020-67699-8},
}
Figure