摘要目的:比较二代测序（NGS）检测方法在痰标本和结核分枝杆菌菌株中耐药性的诊断效能，从而探讨NGS方法检测痰标本耐药性的可行性。方法:本研究属于回顾性研究，收集了北京胸科医院2017年1—12月收治50例肺结核患者的痰标本和相应的临床分离菌株。用NGS方法分别检测痰标本和相应的临床分离菌株 katG、 inhA、 rpoB、 embA、 embB、 rpsL、 rrs、 gyrA、 gyrB、 tlyA的基因突变。采用比例法对菌株进行表型药敏试验（drug susceptibility test，DST）。以DST结果为参考，分别计算临床菌株NGS方法、痰标本NGS方法敏感度、特异度、阳性预测值和阴性预测值，以及与表型DST的一致性统计量（ Kappa）。对于NGS检测在痰标本和菌株样本中的准确性比较用 χ2检验。 结果:在各种突变基因中， rpoB（63.83%，30/47）与 rrs（57.45%，27/47）最常见，其次是 katG（46.81%，22/47）、 rpsL（29.79%，14/47）、 gyrA（27.66%，13/47）、 embB（21.28%，10/47）、 tlyA（12.77%，6/47）、 gyrB（8.51%，4/47），以及 inhA启动子区（19.15%，9/47）、 embA启动子区（12.77%，6/47）突变。NGS方法与异烟肼、利福平、乙胺丁醇、二线注射类药物（链霉素、卷曲霉素、卡那霉素、阿米卡星）、左氧氟沙星的耐药表型相比，在痰样本中，敏感度分别为85.71%、91.67%、77.78%、81.82%、100.00%、87.50%、100.00%、69.23%，特异度分别为100.00%、94.12%、87.50%、89.47%、97.06%、96.97%、94.29%、89.29%。在菌株样本中，敏感度分别为92.86%、100.00%、81.82%、86.96%、88.89%、80.00%、100.00%、85.71%，特异度分别为100.00%、92.86%、87.10%、94.74%、100.00%、100.00%、97.14%、92.86%。与表型药敏结果相比，在痰标本中，NGS方法对异烟肼、利福平、卷曲霉素、卡那霉素、阿米卡星具有较好的检测效能（ Kappa值≥0.75）；在菌株中，NGS方法对异烟肼、利福平、链霉素、卷曲霉素、卡那霉素，阿米卡星及左氧氟沙星具有较好的检测效能（ Kappa值≥0.75）。以NGS方法检测菌株的准确度为参照，检测的所有药物耐药情况在菌株和痰标本中的准确性差异无统计学意义（ P>0.05）。 结论:本研究表明通过NGS技术检测痰标本及菌株基因型来预测异烟肼、利福平及二线注射类药物（卷曲霉素、卡那霉素、阿米卡星）耐药均具有较好的效能，提示NGS方法直接检测痰标本作为早期耐药检测方法的可行性及潜力。

abstractsObjective:To compare the diagnostic performance of next-generation sequencing (NGS) detection methods in sputum samples and Mycobacterium tuberculosis strains, in order to explore the feasibility of the NGS method to detect drug resistance in sputum specimens. Methods:In this retrospective study, the sputum specimens and corresponding clinical isolates of 50 pulmonary tuberculosis patients admitted to Beijing Chest Hospital from January 2017 to December 2017 were collected. The gene mutations of katG, inhA, rpoB, embA, embB, rpsL, rrs, gyrA, gyrB and tlyA in sputum specimens and corresponding clinical isolates were detected by NGS method. The phenotypic drug susceptibility test (DST) of the strains was carried out by the proportion method. Using DST results as a reference, the sensitivity, specificity, positive predictive value and negative predictive value of the NGS method for clinical strains and sputum specimens, as well as the consistency statistic ( Kappa) with phenotype DST were calculated respectively. The Chi-square test was used to compare the accuracy of the NGS testing in sputum samples and strain samples. Results:The results showed that rpoB(63.83%, 30/47) and rrs(57.45%, 27/47) were the most common mutated genes, followed by katG(46.81%, 22/47) , rpsL(29.79%, 14/47), gyrA(27.66%, 13/47), embB(21.28%, 10/47), tlyA(12.77%, 6/47), gyrB(8.51%, 4/47), and inhA promoter(19.15%, 9/47), embA promoter region (12.77%, 6/47) mutation. when the NGS method was compared with the resistance phenotype of isoniazid, rifampicin, ethambutol, second-line injectable drugs (streptomycin, capreomycin, kanamycin, amikacin), levofloxacin, the sensitivity were 85.71%, 91.67%, 77.78%, 81.82%, 100.00%, 87.50%, 100.00%, 69.23%, and the specificity were 100.00%, 94.12, 87.50%, 89.47%, 97.06%, 96.97%, 94.29%, 89.29% in sputum samples, while in strain samples, the sensitivity were 92.86%, 100.00%, 81.82%, 86.96%, 88.89%, 80.00%, 100.00%, 85.71%. The specificity were 100.00%, 92.86%, 87.10%, 94.74%, 100.00%, 100.00%, 97.14%, 92.86%. Compared with the phenotypic drug susceptibility results, the NGS method has better detection performance for isoniazid, rifampicin, capreomycin, kanamycin, and amikacin in sputum specimens ( Kappa≥0.75); while among the strains, the NGS method had a good detection performance for isoniazid, rifampicin, streptomycin, capreomycin, kanamycin, amikacin and levofloxacin ( Kappa≥0.75). With the accuracy of the NGS method for detecting strains as a reference, there was no statistically significant difference in the accuracy of all drug resistance detected between strains and sputum specimens. Conclusions:This study showed that the NGS technology was effective in predicting the resistance of isoniazid, rifampicin, and second-line injectable drugs (capreomycin, kanamycin and amikacin) by detecting sputum samples and strain genotypes, suggesting the feasibility and potential of direct detection of sputum samples by the NGS method as an early detection method for drug resistance.