Abstract:
Both healthy and diseased human ocular surfaces possess their own microbiota. If allowed, opportunistic pathogens within the ocular microbiota may cause microbial keratitis (MK). However, the nonpathogenic component of the ocular microbiota has been proven to undermine the performance of culture, the gold standard of the etiological diagnosis for MK. As the conjunctival bacterial microbiota generates unique alterations with various oculopathies, this study aimed to evaluate the feasibility of distinguishing MK using machine learning based on the characteristics of the conjunctival bacterial microbiome associated with various types of MK. This study also aimed to reveal which bacterial genera constitute the core of the interaction network of the conjunctival bacterial microbiome. Conjunctival swabs collected from the diseased eyes of MK patients and the randomly chosen normal eyes of healthy volunteers were subjected for high-throughput 16S rDNA sequencing. The relative content of each bacterial genus and the composition of bacterial gene functions in every sample were used to establish identification models with the random forest algorithm. Tenfold cross validation was adopted. Accuracy was 96.25% using the bacterial microbiota structure and 93.75% using the bacterial gene functional composition. Therefore, machine learning with the conjunctival bacterial microbiome characteristics might be used for differentiation of MKs as a noninvasive supplementary approach. In addition, this study found that Actinobacteria, Lactobacillus, Clostridium, Helicobacter, and Sphingomonas constitute the core of the interaction network of the conjunctival bacterial microbiome.
摘要:
健康和患病的人眼表都拥有自己的微生物群。如果允许,眼部微生物群内的机会性病原体可能导致微生物性角膜炎(MK)。然而,眼部微生物群的非致病性成分已被证明会破坏培养的性能,MK病因诊断的金标准。由于结膜细菌微生物群产生各种眼病的独特改变,这项研究旨在评估基于与各种类型MK相关的结膜细菌微生物组的特征,使用机器学习区分MK的可行性。这项研究还旨在揭示哪些细菌属构成结膜细菌微生物组相互作用网络的核心。对从MK患者的患病眼睛和随机选择的健康志愿者的正常眼睛收集的结膜拭子进行高通量16SrDNA测序。利用随机森林算法,根据每个样本中各细菌属的相对含量和细菌基因功能组成建立鉴定模型。采用十折交叉验证。使用细菌微生物群结构的准确度为96.25%,使用细菌基因功能组成的准确度为93.75%。因此,具有结膜细菌微生物组特征的机器学习可用于MK的分化,作为一种非侵入性的补充方法。此外,这项研究发现放线菌,乳酸菌,梭菌属,螺杆菌,和鞘氨醇单胞菌构成结膜细菌微生物组相互作用网络的核心。