Skip to content

S-LDSC ALZ Analysis

I applied Stratified Linkage Disequilibrium Score Regression (S-LDSC)1 to summary statistics from Bellenguez et al.'s 2022 Meta-GWAS of Alzheimers' disease2.

Reference Data Sources

I used the standard reference datasets recommended and preprocessed by the authors of the S-LDSC method.

Results

GTEx and Franke lab tissue expression data

Surprisingly, no cell or tissue types were significant at a FDR of 0.01 using GTEx/Franke lab reference dataset.

Here are the cell/tissue types with the lowest p values:

Name Coefficient Coefficient_P_value Reject Null
A15.382.680.Phagocytes 1.07416e-08 0.000303268 False
A15.378.316.Bone.Marrow.Cells 1.04343e-08 0.000613866 False
A15.378.316.580.Monocytes 1.03343e-08 0.000770997 False
A15.382.Immune.System 5.30068e-09 0.0016483 False
A15.382.812.Mononuclear.Phagocyte.System 9.23168e-09 0.00187043 False
A15.382.490.315.583.Neutrophils 5.96328e-09 0.00207589 False
Brain_Substantia_nigra 2.57256e-09 0.00413337 False
Brain_Hippocampus 2.46114e-09 0.0054518 False

The non-significant cell types with the lowest p values are immune related, consistent with theories that immune processes drive Alzheimer's disease (see pg. 1205 of Kandel et al.3).

Roadmap Chromatin data

I next used reference data generated by Finucane et al.1 from the Roadmap Epigenetics Project.

The following graph and table show the results:

s-ldsc-roadmap-Alzheimer

Name Coefficient Coefficient_P_value Reject Null
Primary_B_cells_from_cord_blood__H3K4me1 6.38583e-08 9.12989e-06 True
Spleen_ENTEX__H3K4me3 7.63489e-08 0.000119307 False
Spleen_ENTEX__H3K27ac 5.18872e-08 0.000207013 False
Primary_monocytes_from_peripheral_blood__DNase 1.88703e-07 0.000219091 False
Primary_monocytes_from_peripheral_blood__H3K4me1 6.35288e-08 0.000404116 False
Primary_neutrophils_from_peripheral_blood__H3K4me3 9.90636e-08 0.000756243 False
Spleen_ENTEX__H3K4me1 1.35523e-07 0.000863159 False
Primary_B_cells_from_peripheral_blood__H3K4me1 3.53291e-08 0.000932097 False

In this dataset, there is one significant cell type: B-cells from cord blood. Again, this is consistent with theories of an immune-related etiology3.

ImmGen data

The result of applying S-LSDC to the Alzheimer's GWAS with the ImmGen reference dataset are shown below:

alz-immgen-s-lsdc

Name Coefficient Coefficient_P_value Reject Null
DC.103-11b+F4_80lo.Kd 9.23301e-09 8.0773e-06 True
MF.Microglia.CNS 9.62855e-09 0.000242887 False
Mo.6C-IIint.Bl 1.1929e-08 0.000415892 False
Mo.6C+II+.Bl 9.99718e-09 0.000578951 False
DC.103+11b-.Lu 5.33801e-09 0.00102154 False
NK.49H-.Sp 5.73474e-09 0.00114727 False
GN.BM 5.7906e-09 0.00151902 False
GN.Bl.v2 6.00743e-09 0.00160889 False

Only a single cell type is significant. This cell type appears to be a form of kidney-based dendritic cell.

Corces et al. ATAC-seq data

Application of S-LDSC using the Corces ATAC-seq dataset produces one significant hit. The cell types with the lowest p values are shown below.

Name Coefficient Coefficient_P_value Reject Null
Mono 1.92887e-07 6.37581e-05 True
GMP 3.6411e-08 0.0766454 False
Bcell 3.72995e-08 0.0892083 False
MPP 1.92852e-08 0.19055 False
HSC 1.68987e-08 0.191926 False
MEP 2.10708e-08 0.214319 False
CMP 1.52784e-08 0.259065 False
CLP 2.07019e-08 0.282973 False

It is interesting that the significant hit is the "monocyte" cell type.

Cahoy and GTEx-Brain data

The cell classes with the lowest p values from the Cahoy and GTEx brain datasets are shown below:

Name Coefficient Coefficient_P_value Reject Null
Oligodendrocyte 2.08857e-09 0.154461 False
Astrocyte 4.56992e-10 0.372866 False
Neuron 1.05632e-10 0.467793 False
Name Coefficient Coefficient_P_value Reject Null
Brain_Substantia_nigra 5.26982e-09 0.00104795 False
Brain_Cortex 3.42421e-09 0.0226111 False
Brain_Spinal_cord_(cervical_c-1) 3.25361e-09 0.0239112 False
Brain_Hippocampus 1.31903e-09 0.14029 False
Brain_Nucleus_accumbens_(basal_ganglia) 1.69999e-09 0.170507 False
Brain_Amygdala 9.48447e-10 0.202438 False
Brain_Anterior_cingulate_cortex_(BA24) 1.07874e-09 0.215418 False
Brain_Putamen_(basal_ganglia) 1.23503e-09 0.215505 False

The lack of significant hits is surprising if one views Alzheimer's as a straightforwardly neurological disease. However, these results make sense given theories of Alzheimer's as immune- or microglia- driven.

Comment

It is interesting to contrast the above results with those produced when S-LDSC is applied to inflammatory bowel disease. Both conditions are "immune" in a broad sense, but their S-LDSC results are quite different. Alzheimer's produces a small number of cell-type hits at a moderate level of significance, whereas IBD produces a large number of highly significant cell-type hits. A tentative explanation may be that the immune aspect of Alzheimer's is focused on microglia, and there are no microglia cell types in any of the datasets considered above. Thus, cell types are significant for Alzheimer's only to the extent that they resemble microglia.

  1. Hilary K Finucane, Yakir A Reshef, Verneri Anttila, Kamil Slowikowski, Alexander Gusev, Andrea Byrnes, Steven Gazal, Po-Ru Loh, Caleb Lareau, Noam Shoresh, and others. Heritability enrichment of specifically expressed genes identifies disease-relevant tissues and cell types. Nature Genetics, 50(4):621–629, 2018. URL: https://pmc.ncbi.nlm.nih.gov/articles/PMC5896795/. ↩↩

  2. Céline Bellenguez, Fahri Küçükali, Iris E Jansen, Luca Kleineidam, Sonia Moreno-Grau, Najaf Amin, Adam C Naj, Rafael Campos-Martin, Benjamin Grenier-Boley, Victor Andrade, and others. New insights into the genetic etiology of Alzheimer’s disease and related dementias. Nature Genetics, 54(4):412–436, 2022. URL: https://www.nature.com/articles/s41588-022-01024-z. ↩

  3. Eric R. Kandel, John D. Koester, and Steven A. Mack, Sarah H. Siegelbaum. Principles of neural science 6th edition. Elsevier New York, 2021. URL: https://www.amazon.ca/Principles-Neural-Science-Sixth-Kandel-ebook-dp-B08LNXDCS3/dp/B08LNXDCS3. ↩↩