促排卵是多囊卵巢综合征（PCOS）不孕症的一线治疗方案。卵巢对促排卵治疗的排卵应答差被认为与胰岛素抵抗和高雄激素血症相关。在一个包含1000名PCOS不孕妇女（PCOSAct）的前瞻性队列中，我们开展了一项全外显子联合靶向单核苷酸多态性（SNP）测序以及代谢组学研究。在全基因组水平找出与无排卵显著相关的常见变异和罕见突变，并通过机器学习算法构建排卵预测模型。研究发现，ZNF438基因中标记为rs2994652 (p=2.47×10^–8)的常见变异和REC114基因中的一个罕见功能突变（rs182542888，p=5.79×10^–6）与促排卵治疗失败显著相关。携带rs2994652 A等位基因和REC114 p.Val101Leu (rs182542888)的PCOS不孕妇女进行促排卵治疗的总排卵率更低（分别为：比值比 （OR）=1.96，95% 置信区间（CI）[1.55~2.49]；OR=11.52，95% CI [3.08~43.05]），出现排卵的间隔时间更长（平均56.7天vs.49.0天，p<0.001；78.1天vs.68.6天，p=0.014）。对于rs2994652突变者，L-苯丙氨酸水平升高并与胰岛素抵抗稳态模型（HOMA-IR）指数（r=0.22, p=0.05）和空腹血糖（r=0.33, p=0.003）呈正相关；对于rs182542888突变者，花生四烯酸代谢产物水平下降并与升高的抗苗勒管激素（r=－0.51, p=0.01）和总睾酮（r=－0.71, p=0.02）呈负相关。整合基因变异位点、代谢产物及临床特征的联合预测模型可提高对排卵的预测能力[曲线下面积（AUC）=76.7%]。ZNF438基因的一个常见变异和REC114基因的一个罕见功能突变，以及与二者相关的苯丙氨酸和花生四烯酸代谢物改变，与PCOS女性不孕症的促排卵治疗失败相关。

Ovulation induction is a first-line medical treatment for infertility in polycystic ovary syndrome (PCOS). Poor ovulation responses are assumed to be due to insulin resistance and hyperandrogenism. In a prospective cohort (PCOSAct) of 1000 infertile patients with PCOS, whole-exome plus targeted singlenucleotide polymorphism (SNP) sequencing and comprehensive metabolomic profiling were conducted. Significant genome-wide common variants and rare mutations associated with anovulation were identified, and a prediction model was built using machine learning. Common variants in zinc-finger protein 438 gene (ZNF438) indexed by rs2994652 (p = 2.47 × 10^–8) and a rare functional mutation in REC114 (rs182542888, p = 5.79 × 10^–6) were significantly associated with failure of ovulation induction. Women carrying the A allele of rs2994652 and REC114 p.Val101Leu (rs182542888) had lower ovulation (odds ratio (OR) = 1.96, 95% confidence interval (95%CI) = 1.55–2.49; OR = 11.52, 95%CI = 3.08–43.05, respectively) and prolonged time to ovulation (mean = 56.7 versus (vs) 49.0 days, p < 0.001; 78.1 vs 68.6 days, p = 0.014, respectively). L-phenylalanine was found to be increased and correlated with the Homeostatic Model Assessment for Insulin Resistance (HOMA-IR) index (r = 0.22, p = 0.050) and fasting glucose (r = 0.33, p = 0.003) for rs2994652, while arachidonic acid metabolism was found to be decreased and associated with increased anti-Müllerian hormone (AMH; r = –0.51, p = 0.01) and total testosterone (TT; r = –0.71, p = 0.02) for rs182542888. A combined model of genetic variants, metabolites, and clinical features increased the prediction of ovulation (area under the curve (AUC) = 76.7%). Common variants in ZNF438 and rare functional mutations in REC114, associated with phenylalanine and arachidonic acid metabolites, contributed to the failure of infertility treatment in women with PCOS.