Skip to content

S-LDSC Analysis

I applied Stratified Linkage Disequilibrium Score Regression1 (S-LDSC) to summary statistics from Nyeo et al.'s GWAS of blood EBV DNA levels2.

Reference Data Sources

I used the standard reference datasets prepared by the authors of the S-LDSC method1.

Results

GTEx and Franke lab tissue expression data

I first applied S-LDSC using the GTEx/Franke lab dataset as a reference. At a false discovery rate of 0.01, there were no significant cell/tissue types.

Roadmap epigenetic data

I next applied S-LDSC using the reference dataset derived from the Roadmap epigenetic project. The results are in the plot below:

Results of application of S-LDSC to Nyeo et al.'s EBV DNA GWAS using the epigenetics reference dataset. Points are colored according to broad tissue category. Large points correspond to cell/tissue types deemed significant by an application of the Benjamini-Hochberg procedure at an FDR of 0.01.

At a false discovery rate of 0.01, there are a large number of significant cell types, most of which are T-cells or natural killer cells. This is consistent with the idea that the key determinant of levels of EBV DNA is the ability of the immune system to contain a person's EBV infection, keeping it dormant.

There is also one significant digestive system tissue type. This could be noise, or could perhaps reflect the presence of important immune cells in gut tissue.

That LDSC was able to pick up a much stronger signal from a chromatin epigenetics-based reference dataset compared to a gene-expression based reference dataset is interesting. It is unclear why this would be the case. Potentially, it means that EBV infections are controlled by genes whose differential expression levels are difficult to reliably measure.

ImmGen data

Next, I applied S-LDSC using reference data from the ImmGen project.

There were no significant cell types.

The cell types with the lowest p values are given in the table below:

Name Coefficient Coefficient_P_value Reject Null
T.8SP24-.Th 3.082e-09 0.000625936 False
NKT.4-.Lv 2.58537e-09 0.000767295 False
T.4.PLN.BDC 3.20531e-09 0.000847326 False
T.8Nve.MLN 2.45397e-09 0.00110938 False
NKT.4+.Lv 2.11435e-09 0.00114223 False
T.8Nve.LN 2.6488e-09 0.00151794 False
T.4FP3+25+.Sp 2.31043e-09 0.0015654 False
T.4SP69+.Th 2.52046e-09 0.00255811 False
T.4SP24-.Th 2.57318e-09 0.00291993 False
Tgd.vg2-.act.Sp 1.95198e-09 0.0031735 False
T.4Mem.Sp 3.69048e-09 0.0036837 False
Tgd.vg2+.act.Sp 1.70996e-09 0.00368639 False
T.4FP3-.Sp 2.56193e-09 0.00436498 False
T.4.LN.BDC 2.51854e-09 0.00500759 False
T.4.Pa.BDC 3.65641e-09 0.00502588 False
T.8SP69+.Th 2.60463e-09 0.00506605 False
NKT.4-.Sp 1.64369e-09 0.00586899 False
NKT.44+NK1.1+.Th 2.07751e-09 0.00631805 False
NKT.44-NK1.1-.Th 2.1653e-09 0.0063773 False
T.4SP24int.Th 2.19929e-09 0.00648667 False
Tgd.vg2-.Sp 2.03306e-09 0.00652517 False
Tgd.vg2+.Sp 1.86698e-09 0.00765776 False
T.8Mem.Sp 3.51069e-09 0.00781653 False
T.8Mem.Sp.OT1.d100.LisOva 3.96871e-09 0.00816987 False
Tgd.vg5-.act.IEL 3.31278e-09 0.00946079 False

Consistent with the results above, the lowest p values are all in T or natural killer cells.

It is unclear why the signal from this reference dataset is not sufficient to produce significant cell types.

This could be related to:

  • Differences between the mouse and human immune system
  • Problems with measuring gene expression levels of key genes
  • Since ImmGen is an immune dataset, for a cell type to be significant, there must a differential signal in that cell type compared to other immune cell types. If the EBV DNA GWAS instead provides a generalized immune signal, this could result in no significant cell types.

Corces et al. ATAC-seq data

The results of applying S-LDSC using the epigenetic reference data from Corces et al. ATAC-seq analysis of blood cells are shown below.

Name Coefficient Coefficient_P_value Reject Null
CD8 7.49151e-08 6.86179e-07 True
CD4 6.03118e-08 4.32505e-05 True
NK 6.50971e-08 0.000225251 True
Bcell 3.61798e-08 0.000610793 True
Mono 2.97371e-08 0.0150854 False
GMP 1.45912e-08 0.0519185 False
HSC 1.33929e-08 0.0645347 False
LMPP 1.44259e-08 0.0895894 False
MPP 1.05311e-08 0.0980141 False
CLP 1.79604e-08 0.100229 False
CMP 7.3202e-09 0.17619 False
MEP 3.95883e-09 0.308036 False
Erythro 7.45506e-10 0.475241 False

T-cells (CD4 and CD8), natural killer (NK) cells, and B-cells are all significant. Again, this is consistent with an immune mechanism of control of EBV levels. Also, this is yet another line of evidence reinforcing the primacy of T-cells as the key EBV-level-determining cells.

Cahoy and GTEx-Brain data

The next two reference datasets pertain to the nervous system. The results of running S-LDSC with these two datasets are shown below:

Name Coefficient Coefficient_P_value Reject Null
Astrocyte 1.32785e-09 0.0924435 False
Neuron -6.41258e-10 0.828212 False
Oligodendrocyte -1.82847e-09 0.980079 False
Name Coefficient Coefficient_P_value Reject Null
Brain_Spinal_cord_(cervical_c-1) 1.15863e-09 0.0776225 False
Brain_Amygdala 9.11299e-10 0.139974 False
Brain_Frontal_Cortex_(BA9) 9.32179e-10 0.144977 False
Brain_Cerebellar_Hemisphere 8.02938e-10 0.15864 False
Brain_Cerebellum 5.50179e-10 0.237984 False
Brain_Hypothalamus -2.07267e-10 0.690906 False
Brain_Anterior_cingulate_cortex_(BA24) -2.61295e-10 0.704288 False
Brain_Nucleus_accumbens_(basal_ganglia) -3.27695e-10 0.741726 False
Brain_Putamen_(basal_ganglia) -4.7383e-10 0.748924 False
Brain_Substantia_nigra -5.81302e-10 0.859182 False
Brain_Hippocampus -6.91034e-10 0.867084 False
Brain_Cortex -6.85036e-10 0.898975 False
Brain_Caudate_(basal_ganglia) -8.09745e-10 0.914915 False

There are no significant cell types, which is consistent with the determinants of EBV DNA levels being primarily non-neurological.

  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. Sherry S Nyeo, Erin M Cumming, Oliver S Burren, Meghana S Pagadala, Jacob C Gutierrez, Thahmina A Ali, Laura C Kida, Yifan Chen, Hoyin Chu, Fengyuan Hu, and others. Population-scale sequencing resolves determinants of persistent ebv dna. Nature, pages 1–9, 2026. URL: https://www.nature.com/articles/s41586-025-10020-2. ↩