ABSTRACT

Background

CFH and HTRA1 are pivotal genes driving increased risk for age-related macular degeneration (AMD) among several populations. Here, we performed a hospital-based case-control study to evaluate the effects of three single nucleotide polymorphisms (SNPs) among Hispanics from Mexico.

Materials and methods

122 cases and 249 controls were genotyped using Taqman probes. Experienced ophthalmologists diagnosed AMD following the American Association of Ophthalmology guidelines. We studied CFH (rs1329428, rs203687) and HTRA1 (rs11200638) SNPs thoroughly by logistic regression models (assuming different modes of inheritance) and machine learning-based methods (ML).

Results

HTRA1 rs11200638 is the most significant polymorphism associated with AMD in our studied population. In a multivariate regression model adjusted for clinically and statistically meaningful covariates, the A/G and A/A genotypes increased the odds of disease by a factor of 2.32 and 7.81, respectively (P < .05) suggesting a multiplicative effect of the polymorphic A allele. Furthermore, this observation remains statistically meaningful in the allelic, dominant, and recessive models, and ML algorithms. When stratifying by phenotype, this polymorphism was significantly associated with increased odds for geographic atrophy (GA) in a recessive mode of inheritance (12.4, p < .05).

Conclusions

In sum, this work supports a strong association between HTRA1 genetic variants and AMD in Hispanics from Mexico, especially with GA. Moreover, ML was able to replicate the results of conventional biostatistics methods unbiasedly.

中文翻译：

---

使用经典和机器学习方法评估年龄相关性黄斑变性中的 CFH 和 HTRA1 多态性。

 抽象的

 背景

CFH和HTRA1是导致多个人群年龄相关性黄斑变性 (AMD) 风险增加的关键基因。在这里，我们进行了一项以医院为基础的病例对照研究，以评估三种单核苷酸多态性 (SNP) 对墨西哥西班牙裔人的影响。

 材料和方法

使用 Taqman 探针对 122 例病例和 249 例对照进行基因分型。经验丰富的眼科医生根据美国眼科协会的指南诊断 AMD。我们通过逻辑回归模型（假设不同的遗传模式）和基于机器学习的方法（ML）彻底研究了CFH （rs1329428，rs203687）和HTRA1 （rs11200638）SNP。

 结果

HTRA1 rs11200638 是我们研究人群中与 AMD 相关的最显着的多态性。在针对临床和统计上有意义的协变量进行调整的多变量回归模型中，A/G 和 A/A 基因型分别将疾病几率增加了 2.32 和 7.81 倍 ( P < .05)，表明多态性 A 具有乘数效应等位基因。此外，这一观察结果在等位基因、显性和隐性模型以及机器学习算法中仍然具有统计意义。当按表型分层时，这种多态性与隐性遗传模式下地理萎缩（GA）几率的增加显着相关（12.4， p < .05）。

 结论

总之，这项工作支持HTRA1基因变异与墨西哥西班牙裔 AMD 之间的密切关联，尤其是与 GA 相关。此外，机器学习能够公正地复制传统生物统计学方法的结果。