State Key Laboratory of Environment Health (Incubation), MOE (Ministry of Education) Key Laboratory of Environment & Health, Ministry of Environmental Protection Key Laboratory of Environment and Health (Wuhan), and Department of Epidemiology and Biostatistics, School of Public Health, Tongji Medical College, Huazhong University of Science and Technology, Wuhan, 430030, China.
Breast Cancer Res Treat. 2013 Dec;142(3):637-44. doi: 10.1007/s10549-013-2775-8. Epub 2013 Nov 22.
Genome-wide association studies (GWASs) of breast cancer (BC) have identified multiple risk variants. However, the multiple interactions among these variants are still not well established. In this study, we utilized the multi-analytic strategy combing random forest (RF), multifactor dimensionality reduction (MDR), and logistic regression approaches to investigate the high-order interactions among ten genetic variants recently identified by GWAS in 477 BC patients and 534 healthy controls. Expectedly, six variants, rs1219648, rs3757318, rs1926657, rs6556756, rs2046210, and rs4973768, were significantly associated with BC risk under independent analysis. In RF analysis, rs3757318, rs2046210, and rs4973768 were ranked as the top three important risk factors and were selected as the best set which taking interactions into consideration. Subsequently, the MDR analysis of the ten variants found that the three-factor model including rs3757318, rs2046210, and rs4973768 interpret the best interaction model with the maximized testing accuracy of 0.6183 and cross-validation consistency of 10/10. Intriguingly, cumulative effect was observed in the manner of dose-dependent with increasing numbers of risk alleles (P(trend) = 9.80 × 10(-5)), and the individuals carrying 4-6 risk alleles had a threefold higher risk of BC than carrying 0 risk alleles (OR 3.27, 95 % CI 1.96-5.48). Our findings emphasized the proof of principle that multiple interactions of genetic variants, including rs3757318, rs2046210, and rs4973768 may play important roles in the susceptibility of BC though the biological mechanisms underlying the observed associations need to be elucidated.
全基因组关联研究（GWAS）已经确定了多个乳腺癌（BC）风险变异。然而，这些变异之间的多重相互作用仍然没有得到很好的建立。在这项研究中，我们利用了结合随机森林（RF）、多因素维度缩减（MDR）和逻辑回归方法的多分析策略，来研究最近在 477 名 BC 患者和 534 名健康对照者的 GWAS 中确定的十个遗传变异之间的高阶相互作用。预期地，rs1219648、rs3757318、rs1926657、rs6556756、rs2046210 和 rs4973768 六个变异在独立分析中与 BC 风险显著相关。在 RF 分析中，rs3757318、rs2046210 和 rs4973768 被列为前三个重要的危险因素，并被选为考虑相互作用的最佳集合。随后，对十个变异的 MDR 分析发现，包括 rs3757318、rs2046210 和 rs4973768 的三因素模型解释了最佳的相互作用模型，其测试精度最大为 0.6183，交叉验证一致性为 10/10。有趣的是，在数量依赖的方式中观察到了累积效应，风险等位基因的数量增加（P(trend) = 9.80 × 10(-5))，携带 4-6 个风险等位基因的个体比携带 0 个风险等位基因的个体患 BC 的风险高三倍（OR 3.27，95 % CI 1.96-5.48）。我们的研究结果强调了一个原则性的证明，即包括 rs3757318、rs2046210 和 rs4973768 在内的遗传变异的多重相互作用可能在 BC 的易感性中发挥重要作用，尽管需要阐明观察到的关联背后的生物学机制。
