DataSheet2_LncRNA model predicts liver cancer drug resistance and validate in vitro experiments.CSV

Introduction: Hepatocellular carcinoma (HCC) patients may benefit from chemotherapy, but drug resistance is an important obstacle to favorable prognoses. Overcoming drug resistance is an urgent problem to be solved.
Methods: Differential expression analysis was used to identify long non-coding RNAs (LncRNAs) that differed in chemotherapy-sensitive and chemotherapy-resistant patients. Machine learning algorithms including random forest (RF), lasso regression (LR), and support vector machines (SVMs) were used to identify important chemotherapy-related LncRNAs. A back propagation (BP) network was then used to validate the predictive capacity of important LncRNAs. The molecular functions of hub LncRNAs were investigated via qRT-PCR and cell proliferation assay. Molecular-docking technique was used to explore candidate drug of targets of hub LncRNA in the model.
Results: A total of 125 differentially expressed LncRNAs between sensitive and resistant patients. Seventeen important LncRNAs were identified via RF, and seven factors were identified via LR. With respect to SVM, the top 15 LncRNAs of AvgRank were selected. Five merge chemotherapy-related LncRNAs were used to predict chemotherapy resistance with high accuracy. CAHM was a hub LncRNA of model and expression high in sorafenib resistance cell lines. In addition, the results of CCK8 showed that the sensitivity of HepG2-sorafenib cells to sorafenib was significantly lower than that of HepG2; and the sensitivity of HepG2-sorafenib cells transfected with sh-CAHM was significantly higher than that of Sorafenib. In the non-transfection group, the results of clone formation experiments showed that the number of clones formed by HepG2-sorafenib cells treated with sorafenib was significantly more than that of HepG2; after HepG2-sorafenib cells were transfected with sh-CAHM, the number of clones formed by Sorafenib treatment was significantly higher than that of HepG2 cells. The number was significantly less than that of HepG2-s + sh-NC group. Molecular Docking results indicate that Moschus was candidate drug for target protein of CAHM.
Conclusion: Five chemotherapy-related LncRNAs could predict drug resistance in HCC with high accuracy, and the hub LncRNA CAHM has potential as a new biomarker for HCC chemotherapy resistance.

引言：肝细胞癌（hepatocellular carcinoma, HCC）患者可从化疗中获益，但化疗耐药是影响良好预后的重要阻碍，克服化疗耐药是亟待解决的临床问题。
方法：本研究通过差异表达分析，筛选化疗敏感与化疗耐药患者之间存在表达差异的长链非编码RNA（long non-coding RNAs, LncRNAs）；采用随机森林（random forest, RF）、套索回归（lasso regression, LR）以及支持向量机（support vector machines, SVMs）等机器学习算法，筛选与化疗相关的关键长链非编码RNA；随后构建反向传播（back propagation, BP）神经网络，验证筛选得到的关键长链非编码RNA的预测效能；通过实时荧光定量PCR（quantitative real-time polymerase chain reaction, qRT-PCR）及细胞增殖实验，探究核心长链非编码RNA的分子功能；采用分子对接技术，筛选模型中核心长链非编码RNA靶标的潜在治疗药物。
结果：本研究在化疗敏感与耐药患者之间共筛选得到125个差异表达的长链非编码RNA。通过随机森林算法筛选得到17个关键长链非编码RNA，套索回归算法筛选得到7个关键因子；针对支持向量机算法，选取平均排名（AvgRank）前15的长链非编码RNA。最终整合得到5个与化疗相关的长链非编码RNA，可高精度预测化疗耐药情况。CAHM是本模型的核心长链非编码RNA，在索拉非尼（sorafenib）耐药细胞系中呈高表达状态。此外，CCK8实验结果显示，HepG2-索拉非尼耐药细胞（HepG2-sorafenib）对索拉非尼的敏感性显著低于亲本HepG2细胞；而转染sh-CAHM的HepG2-sorafenib细胞对索拉非尼的敏感性显著高于未转染组细胞。克隆形成实验结果显示，经索拉非尼处理的HepG2-sorafenib细胞所形成的克隆数显著多于亲本HepG2细胞；而转染sh-CAHM的HepG2-sorafenib细胞经索拉非尼处理后，其克隆形成数显著高于亲本HepG2细胞，但显著低于HepG2-s + sh-NC组。分子对接结果显示，麝香（Moschus）可作为CAHM靶标蛋白的潜在治疗药物。
结论：5个与化疗相关的长链非编码RNA可高精度预测肝细胞癌的化疗耐药情况，且核心长链非编码RNA CAHM有望成为肝细胞癌化疗耐药的新型生物标志物。