Archives

  • 2019-07
  • 2019-08
  • 2019-09
  • 2019-10
  • 2019-11
  • 2020-03
  • 2020-07
  • 2020-08
  • mLOY and Lung Cancer Risk

    2020-08-14

     mLOY and Lung Cancer Risk in Chinese 39
    chromosome: 2,694,521–59,034,049, hg19/GRCh37), in 2255 males (1445 case patients and 810 controls) from the aforementioned 3797 male samples with available raw array data. The signal intensity, genotype call, and confidence files derived from the Affymetrix Power Tools software were further analyzed with the PennCNV-Affy pipeline software and produced the LRR for each probe. Assuming that the total variation in mLRR-Y consisted of signals from both mLOY and experimental factors, we further excluded the influence of these factors as previously described.4 We first determined the peak of the histogram of mLRR-Y with a kernel density estimation method implemented in the density function of R and use of the bandwidth SJ. Then, the bin with the highest value in the smoothed distribution was selected as the local me-dian, and the observations in the positive tail were mirrored over the kernel-derived median to create the negative tail. Finally, for the most conservative estimate of the frequency of mLOY, the lowest value in the simulated noise distribution was used as the threshold.4
    Statistical Analysis
    We created a wGRS to predict mLOY for mendelian randomization by summing the dosage of 14 indepen-dent mLOY-increasing 454-29-5 with the following formula:
    the i-th SNP for mLOY from the previous study5 and SNPi is the dosage (coded as 0, 1, or 2 for wild-type homo-zygous, heterozygous, or homozygous, respectively) of the effect allele which is associated with increasing mLOY.
    We performed linear regression to evaluate the asso-ciation between mLRR-Y and each of the 14 SNPs in the controls, adjusting for age and smoking status. A restricted cubic spline function was applied to examine the shape of the association of the genetically predicted mLOY and mLRR-Y with lung cancer risk, and the Wald test was used for the nonlinearity estimation. The association between genetically predicted mLOY and lung cancer risk was estimated by logistic regression analysis, assuming an additive effect of the allele dosage on the log odds scale adjusted for age, pack-years of smoking, and the first PC. Stratification analysis was performed by known factors that could influence individual mLOY levels, including age and smoking status. The Cochran Q statistic was calcu-lated to test for heterogeneity in two subgroups.
    To determine the robustness of the association between genetically predicted mLOY and lung cancer risk, we performed sensitivity analysis with both the unweighted genetic risk score (GRS) and the
    inverse variance–weighted method. The GRS was calcu-lated by summing the dosage of 14 independent mLOY-increasing alleles with the following formula: GRS ¼ 14 P SNPi. The inverse variance–weighted regression was
    applied to examine the potential causal association be-tween mLOY (X) and lung cancer risk (Y) with mLOY-related SNPs as the instrumental variables.11 Briefly, the causal effect (bYX) between mLOY and lung cancer risk was evaluated with the Wald estimator: bYX ¼ bYG , where bYG is the estimated effect bXG
    for lung cancer associated with the instrumental variable and bXG is the estimated effect for mLOY associated with the instrumental variable obtained from the previous GWAS.5 Although the SE for mLOY associated with the instrumental variable was not provided in the reported GWAS, we calculated SXG with the following for- mula 12
    bXG
    the causal effect was ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi the delta
    calculated by using
    rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiðÞþ
    method11: SE YX
    SYG 2 ðSXG bYG Þ2 , where S and
    b XG b 4
    XG
    XG
    SYG are the corresponding SEs. We also performed mendelian randomization–Egger regression analysis to estimate the potential pleiotropic effect of the mLOY-related genetic variants.13 A multivariate Cox propor-tional hazards regression model was used to estimate the hazard ratios (HRs) and their 95% confidence intervals (CIs), adjusting for age, pack-years of smoking, and tumor stage. General statistical analyses were performed with R software (version 3.2.2). Two-sided p values less than 0.05 were considered statistically significant.
    Results
    Genetically Predicted mLOY Is Associated with mLRR-Y
    A total of 14 mLOY-related SNPs were included in the analysis after exclusion of five variants with MAF less than 0.05 in the Chinese population. We created a wGRS to predict mLOY by summing the dosage of 14 inde-pendent mLOY-increasing alleles (see Supplementary Table 2).