Archives

  • 2019-07
  • 2019-08
  • 2019-09
  • 2019-10
  • 2019-11
  • 2020-03
  • 2020-07
  • 2020-08
  • br We first estimated mLRR Y as a proxy of

    2020-08-14


    We first estimated mLRR-Y as a proxy of the copy number of the Y chromosome in 810 healthy male participants; the association between 14 reported mLOY-related SNPs and mLRR-Y is shown in Supplementary Table 3. A significant negative correla-tion between genetically predicted mLOY and mLRR-Y was observed, whereas individuals carrying greater numbers of mLOY-increasing 30931-67-0 had a significantly lower mLRR-Y (b ¼ –0.58, p ¼ 3.38 10 2) (Fig. 1 and Supplementary Fig. 1), suggesting that the predicted mLOY from the 14 mLRR-Y signals identified in Wright 
    et al.5 is a good indicator of the individual quantitative level of Y chromosome loss. Additionally, consistent with previous results, the individual mLRR-Y level was significantly greater in lifelong nonsmokers than in smokers (p ¼ 1.52 10–2) and the younger population (p ¼ 3.28 10–5) (Supplementary Fig. 2A and B).
    Genetically Predicted mLOY Is Associated with a Decreased Lung Cancer Risk
    We provide the 14 mLOY-related genetic variants and their effect on lung cancer risk from our lung cancer GWAS data in Supplementary Table 3. We performed a logistic regression analysis to evaluate the association between genetically predicted mLOY and lung cancer risk in 3797 males (1711 case patients and 2086 con-trols). After controlling for age, smoking status, and the first PC, we found that increased genetically predicted mLOY was significantly associated with a decreased risk of lung cancer (per SD increase in the wGRS: OR ¼ 0.92, 95% CI: 0.86–0.98, p ¼ 1.30 10–2) (see Fig. 1 and Supplementary Fig. 3). The more interesting result was that the protective effect was observed in lifelong nonsmokers but not in smokers (lifelong nonsmokers: OR ¼ 0.80, 95% CI: 0.69–0.93, p ¼ 4.03 10–3; smokers: OR ¼ 0.96, 95% CI: 0.89–1.04, p ¼ 2.90 10–1; pHeterogeneity ¼ 3.83 10–2) (Table 1), suggesting that smoking behavior dramatically abolished the protective effect of mLOY against lung cancer.
    Because smoking behavior and increasing age contribute to the generation of mLOY and the risk of lung cancer, we further investigated whether mLOY caused by damaging environmental factors was different from genetically defined mLOY. As expected, we observed that genetically predicted mLOY was not associated with smoking status or age (Supplementary Fig. 2C and D), whereas the individual 30931-67-0 mLRR-Y level was significantly greater in lifelong nonsmokers and the younger popu-lation than in smokers and the older population (see Supplementary Fig. 2A and B). More interestingly, whereas genetically predicted mLOY was linearly asso-ciated with a decreased lung cancer risk (p for linearity
    Table 1. Stratified Analyses of the Association between mLOY and Lung Cancer Risk
    Male
    Age/Smoking Status OR (95% CI) p Value Q Value
    Age
    Smoke status
    mLOY, mosaic loss of chromosome Y; CI, confidence interval.
    January 2019 mLOY and Lung Cancer Risk in Chinese 41
    Genetically Predicted mLOY Is Associated with a Better Lung Cancer Prognosis
    To further evaluate the effect of mLOY on lung cancer prognosis, we performed a multivariate Cox proportional hazard regression analysis in 309 male patients with lung cancer treated without surgery. Interestingly, we found that increased genetically mLOY was significantly associated with a better prognosis after controlling for age, smoking status, and tumor stage (crude HR ¼ 0.87, 
    Sensitivity Analysis Using Alternative Causal Inference Methods
    We further evaluated the association between 14 SNPs and lung cancer risk in our case-control study to
    A
    Odds
    Log
    − Weighted genetic risk score
    Odds
    Log
    mLRR of Chromosome Y
    Figure 2. Association between mosaic loss of chromosome Y and lung cancer risk based on the restricted cubic spline function in the logistic regression model. (A) Linear association between genetically predicted mosaic loss of chromosome Y and lung cancer risk. (B) U-shaped association between median of the log R ratio (mLRR) of probes in the Y chromosome and lung cancer risk.
    Genetically predicted low mLOY
    Genetically predicted high mLOY
    Month
    Figure 3. Kaplan-Meier estimator of genetically predicted mosaic loss of chromosome Y (mLOY) in lung cancer samples. Abbreviations: HR, hazard ratio; CI, confidence interval.
    infer whether any of the individual genetic variants used in our wGRS drove this result (Supplementary Table 3); however, none of these SNPs had significant associations with lung cancer risk, and no heterogeneity was observed (p for a Cochran Q test value of 0.76). The aforementioned result suggested that the effect esti-mated from the genetically predicted mLOY was not driven by any single SNP.
    To test the robustness of our primary result, we further evaluated the association between genetically predicted mLOY and lung cancer by using several addi-tional methods (see Supplementary Fig. 3). First, we constructed an unweighted risk score with the afore-mentioned 14 SNPs and found a consistent association with the results derived from wGRS (per SD increase: OR ¼ 0.92, 95% CI: 0.86–0.98, p ¼ 1.35 10–2), sug-gesting that the choice of weights used in the wGRS method was robust. Then, we obtained both weighted and unweighted median estimates for the causal effect using all of the aforementioned genetic variants. The causal effect generated by this approach was similar to that generated by the aforementioned GRS approach (per SD increase in unweighted median-based estimate: OR ¼ 0.88, 95% CI: 0.79–0.97, p ¼ 1.00 10–2; per SD increase in weighted median-based estimate: OR ¼ 0.89, 95% CI: 0.81–0.98, p ¼ 2.10 10–2). This median-based estimate method is unbiased asymptotically and has less stringent requirements than the genetic risk score method, which required only at least half of the weight for the score to be derived from genetic instruments.14 Additionally, we also evaluated the potential causal ef-fect with the inverse-variance weighted method, and a similar result was observed (OR ¼ 0.92, 95% CI: 0.86– 0.98, p ¼1.50 10–2).