Rationale: Idiopathic pulmonary fibrosis (IPF) is characterized by progressive lung scarring and death. Early detection is paramount. Chest computed tomographic imaging can detect a precursor phenotype known as interstitial lung abnormalities (ILA). The MUC5B promoter variant rs35705950 is a common variant of large effect size that confers risk for both IPF (OR 6-13, heterozygotes) and ILA (OR 2-3). The large effect size of this variant may mask identification of polygenic effects. We hypothesized that a polygenic risk score (PRS) excluding the MUC5B region would add complementary predictive value to the MUC5B variant for IPF and ILA.

Methods: Using previously published genome-wide association summary statistics for IPF, we used lassosum to develop an IPF PRS excluding the MUC5B region (±250 kb). We trained a composite risk score of the PRS and MUC5B rs35705950 (PRS-M5B) using 10-fold cross-validation (100 bootstraps) in IPF cases and controls from the external Lung Tissue Research Consortium (LTRC). We tested the associations of the PRS, rs35705950, and PRS-M5B with IPF in LTRC and ILA in the Genetic Epidemiology of COPD (COPDGene) study. Multivariable logistic regressions were performed, adjusting for age, sex, smoking, and genetic ancestry. Area-under-the-curve (AUC) analyses were used to assess model predictive performances. We examined phenotypic variance explained by each risk factor using Nagelkerke R2, and also compared this measure to a traditional PRS including the MUC5B region.

Results: In LTRC (270 controls, 255 IPF cases), the PRS (OR 4.5 [95% CI: 3.3 - 6.1], p=8.0e-22) and rs35705950 (OR 3.6 [95% CI: 2.4 - 5.2], p=6.4e-11) were both associated with IPF, and when considered in a single model, effect sizes were largely unchanged. The cross-product interaction term (PRS X rs35705950) was not significant (p=0.7). PRS-M5B demonstrated a larger effect (OR 18 [95% CI: 10 – 30], p=5.0e-26) compared to either component risk factor. There was a stepwise increase in predictive performances of models including rs35705950 (AUC 0.74), PRS (AUC 0.83), and PRS-M5B (AUC 0.86). In COPDGene (6417 controls, 250 ILA), rs35705950 (OR 1.8 [95% CI: 1.4-2.4], p=2.7e-6), but not the PRS (p=0.08), were significantly associated with ILA. Phenotypic variability in IPF explained by PRS-M5B (R2=0.41) was greater than PRSs including (R2=0.34) and excluding (R2=0.33) the MUC5B region, or rs35705950 alone (R2=0.12).

Conclusions: The MUC5B rs35705950 variant and a PRS excluding the MUC5B region offered complementary predictive value for IPF, but not ILA. For certain traits, such as IPF, separating out common variants of large effect size may aid in identifying polygenic effects.