Linkage Disequilibrium Score Regression for 33 Chemokines

---

1 Curse of Knowledge

2 Syllables

Syllables

An-ti-dis-es-tab-lish-men-tar-i-an-is-m

3 LD Score Regression

LDSC

LD Score Regression (LDSC)

---

4 Equation

Equation

\[ \text{E}[\chi^2] = 1 + N \cdot h^2 \cdot \text{LD Score} + \text{Intercept} \]

---

Show

4.1 E[χ²]: Expected Chi-Squared Statistic

• Derived from GWAS summary statistics:
  • \(\chi^2 = \left(\frac{\beta}{\text{SE}}\right)^2\), where:
    • \(\beta\): The estimated effect size of the SNP.
    • \(\text{SE}\): The standard error of the effect size estimate.
  • Larger \(\chi^2\) values indicate stronger evidence of an association between the SNP and the trait.

---

Show

4.2 1: Random Noise

• Accounts for baseline variance or “statistical noise” in the data.

---

Show

4.3 N: Sample Size

• number of individuals included in the GWAS.
• LDSC best powered for N>5000

---

Show

4.4 h²: SNP-Based Heritability

• fraction of the trait variance explained by all genotyped SNPs

---

Show

4.5 LD Score

• measure of how much a SNP is correlated with nearby SNPs.

The LD Score for a SNP \(j\) is calculated as:

\[ \text{LD Score for SNP } j = \sum_{i} r^2_{ij} \]

Where: - \(r^2_{ij}\):

The squared correlation (linkage disequilibrium) between SNP \(j\) and SNP \(i\).

---

Show

4.6 Intercept: Bias Adjustment

• Adjusts for confounding factors such as:
  • ancestry, relatedness, sample overlap

---

5 Post-GWAS Method

Post-GWAS

• Ingests GWAS summary statistics as its data
• Like two-sample Mendelian randomization (MR) and colocalization
• Uses the pattern of association between SNPs
• Implemented originally in Python for SNP-heritability and genetic correlations for pairs of traits
• Implemented in GenomicSEM, its products (matrices) can be used for further modeling to produce 💡NEW💡 GWAS summary statistics

Case-control study of blindness in the immune-compromised

IL6

Interleukin 6

Mini-Cytokines

Molecule Type	Molecular Weight (kDa)	Examples
Cytokines	~ 15–30 kDa	IL-6 (~21 kDa), TNF-α (~17 kDa), IFN-γ (~20–25 kDa)
Chemokines	~ 8–12 kDa	CXCL8 (IL-8, ~8 kDa), CCL2 (MCP-1, ~11 kDa), CX3CL1 (~10 kDa)

6 LDSC of 33 Chemokines

6.1 UK Biobank Pharma Proteomics Project (UKB-PPP)

• UKB-PPP is a collaboration between the UK Biobank (UKB) and 13 biopharmaceutical companies

• Produced GWAS summary statistics for Olink proteins

• N ~ 30,000K

7 Results

7.1 Heatmap

Hierarchical Clustering Heatmap for Genetic Correlations between 33 Olink Chemokines

7.2 Table