BACKGROUND & AIMS Relatives of individuals with Crohn's disease (CD) carry CD-associated genetic variants and are often exposed to environmental factors that increase their risk for this disease. We aimed to estimate the utility of genotype, smoking status, family history, and other biomarkers can be used to calculate risk in asymptomatic first-degree relatives of patients with CD. METHODS We recruited 480 healthy first-degree relatives (full siblings, offspring or parents) of patients with CD through the Guy's and St Thomas' NHS Foundation Trust and from members of Crohn's and Colitis, United Kingdom. DNA samples were genotyped using the Immunochip. We calculated a risk score for 454 participants, based on 72 genetic variants associated with CD, family history, and smoking history. Participants were assigned to highest and lowest risk score quartiles. We assessed pre-symptomatic inflammation by capsule endoscopy and measured 22 markers of inflammation in stool and serum samples (reference standard). Two machine-learning classifiers (elastic net and random forest) were used to assess the ability of the risk factors and biomarkers to identify participants with small intestinal inflammation in the same dataset. RESULTS The machine-learning classifiers identified participants with pre-symptomatic intestinal inflammation: elastic net (area under the curve, 0.80; 95% CI, 0.62-0.98) and random forest (area under the curve, 0.87; 95% CI, 0.75-1.00). The elastic net method identified 3 variables that can be used to calculate odds for intestinal inflammation: combined family history of CD (odds ratio, 1.31), genetic risk score (odds ratio, 1.14), and fecal level of calprotectin (odds ratio, 1.04). These same 3 variables were among the 5 factors associated with intestinal inflammation in the random forest model. CONCLUSION Using machine learning classifiers, we found that genetic variants associated with CD, family history, and fecal level of calprotectin together identify individuals with pre-symptomatic intestinal inflammation who are therefore at risk for CD. A tool for detecting people at risk for CD before they develop symptoms would help identify the individuals most likely to benefit from early intervention.

中文翻译：

---

遗传和炎症生物标志物对克罗恩病患者无症状一级亲属的小肠炎症进行分类。

背景和目的 克罗恩病 (CD) 患者的亲属携带 CD 相关的遗传变异，并且经常暴露于环境因素，这会增加他们患这种疾病的风险。我们旨在评估基因型、吸烟状况、家族史和其他生物标志物的效用，这些生物标志物可用于计算 CD 患者无症状一级亲属的风险。方法 我们通过 Guy 和 St Thomas 的 NHS 信托基金会以及英国克罗恩病和结肠炎的成员招募了 480 名 CD 患者的健康一级亲属（兄弟姐妹、后代或父母）。使用免疫芯片对 DNA 样品进行基因分型。我们根据与 CD、家族史和吸烟史相关的 72 种遗传变异计算了 454 名参与者的风险评分。参与者被分配到最高和最低风险评分四分位数。我们通过胶囊内窥镜评估了症状前炎症，并测量了粪便和血清样本（参考标准）中的 22 种炎症标志物。两个机器学习分类器（弹性网络和随机森林）用于评估风险因素和生物标志物在同一数据集中识别患有小肠炎症的参与者的能力。结果 机器学习分类器识别出有症状前肠道炎症的参与者：弹性网（曲线下面积，0.80；95% CI，0.62-0.98）和随机森林（曲线下面积，0.87；95% CI，0.75- 1.00）。弹性网法确定了 3 个可用于计算肠道炎症几率的变量：CD 的综合家族史（优势比，1.31），遗传风险评分（优势比，1.14）和粪便钙卫蛋白水平（优势比，1.04）。这些相同的 3 个变量是随机森林模型中与肠道炎症相关的 5 个因素之一。结论 使用机器学习分类器，我们发现与 CD、家族史和粪便钙卫蛋白水平相关的遗传变异共同识别出有症状前肠道炎症的个体，这些个体因此有患 CD 的风险。在出现症状之前检测有 CD 风险的人的工具将有助于确定最有可能从早期干预中受益的个人。我们发现与 CD、家族史和粪便钙卫蛋白水平相关的遗传变异共同确定了有症状前肠道炎症的个体，因此他们有患 CD 的风险。在出现症状之前检测有 CD 风险的人的工具将有助于确定最有可能从早期干预中受益的个人。我们发现与 CD、家族史和粪便钙卫蛋白水平相关的遗传变异共同确定了有症状前肠道炎症的个体，因此他们有患 CD 的风险。在出现症状之前检测有 CD 风险的人的工具将有助于确定最有可能从早期干预中受益的个人。