基于深度卷积生成对抗网络算法的肺炎细菌sRNA识别及潜在的药物靶标研究

Pneumonia is a common infectious disease and the leading cause among infection-related death of children worldwide. The pathogens of the disease mainly include bacteria such as Streptococcus pneumoniae, Klebsiella pneumoniae, Mycoplasma pneumoniae and Chlamydia pneumoniae. As a kind of non-coding RNA unique to bacteria, sRNA is a very suitable potential drug target. Thus, it showed great practical significance for the treatment of pneumonia to identify sRNA in pneumonia-related bacteria. This project proposes to use deep convolutional generative adversarial network algorithm to identify sRNA in bacterial genome, which can effectively solve the small-sample problem encountered in sRNA recognition. We applied this method into four categories of pneumonia-related bacterial genomes to find out the sRNA transcribed from these bacteria. In addition, we intend to screen out potential sRNA drug targets by identifying target genes for sRNA, analyzing regulatory networks, and distinguishing essential genes. The implementation of the project will not only provide an effective theoretical basis but therapeutic clues for the treatment of pneumonia.

肺炎是一种常见的感染性疾病，也是全世界儿童因感染致死的主要原因。该疾病的病原菌主要包括肺炎链球菌、肺炎克雷伯杆菌、肺炎支原体和肺炎衣原体等细菌。sRNA作为细菌特有的一类非编码RNA，是一种非常合适的药物靶标。因此，识别肺炎细菌sRNA对于治疗肺炎有着重大的现实意义。本项目提出以深度卷积生成对抗网络算法，来识别细菌基因组中的sRNA，能够有效地解决sRNA识别所遇到的小样本问题。我们把方法进一步应用到四种常见的肺炎细菌基因组中，找出这些细菌所转录的sRNA。另外，我们拟通过识别sRNA的靶基因、分析调控网络和鉴定必需基因等方法，来筛选出潜在的sRNA药物靶标。该项目的实施可为肺炎疾病的治疗提供有效的理论依据和治疗线索。

肺炎是一种常见的感染性疾病，也是全世界儿童因感染致死的主要原因。该疾病的病原菌主要包括肺炎链球菌、肺炎克雷伯杆菌和肺炎支原体等细菌。sRNA作为细菌特有的一类非编码RNA，非常适合作为一个切入点来研究细菌的药物靶标。本项目采用了深度卷积生成对抗网络算法，来识别细菌基因组中的sRNA，构建了sRNA的预测模型，通过绘制ROC曲线，其预测总体精度达到了0.872。我们把模型应用于结核分枝杆菌中，找到了该细菌中的16个sRNA。通过生物信息挖掘，我们找到了结核分枝杆菌中的12个毒力因子和6个药物靶标。同时，我们也在支原体中，找到了19个sRNA，8个毒力因子，两个可靠的药物靶标(nadD和pheS)。我们的结果为这两种细菌感染疾病的预防和治疗提供了有效的理论依据和治疗线索。