Luo Yue-Mei, Liu Fei-Tong, Chen Mu-Xuan, Tang Wen-Li, Yang Yue-Lian, Tan Xi-Lan, Zhou Hong-Wei
Department of Environmental Health, School of Public Health, Southern Medical University, Guangzhou 510515, China. E-mail:
Nan Fang Yi Ke Da Xue Xue Bao. 2018 Mar 20;38(3):251-260. doi: 10.3969/j.issn.1673-4254.2018.03.03.
To investigate the effects of prebiotics supplementation for 9 days on gut microbiota structure and function and establish a machine learning model based on the initial gut microbiota data for predicting the variation of Bifidobacterium after prebiotic intake.
With a randomized double-blind self-controlled design, 35 healthy volunteers were asked to consume fructo-oligosaccharides (FOS) or galacto-oligosaccharides (GOS) for 9 days (16 g per day). 16S rRNA gene high-throughput sequencing was performed to investigate the changes of gut microbiota after prebiotics intake. PICRUSt was used to infer the differences between the functional modules of the bacterial communities. Random forest model based on the initial gut microbiota data was used to identify the changes in Bifidobacterium after 5 days of prebiotic intake and then to build a continuous index to predict the changes of Bifidobacterium. The data of fecal samples collected after 9 days of GOS intervention were used to validate the model.
Fecal samples analysis with QIIME revealed that FOS intervention for 5 days reduced the intestinal flora alpha diversity, which rebounded on day 9; in GOS group, gut microbiota alpha diversity decreased progressively during the intervention. Neither FOS nor GOS supplement caused significant changes in β diversity of gut microbiota. The area under the curve (AUC) of the prediction model was 89.6%. The continuous index could successfully predict the changes in Bifidobacterium (R=0.45, P=0.01), and the prediction accuracy was verified by the validation model (R=0.62, P=0.01).
Short-term prebiotics intervention can significantly decrease α-diversity of the intestinal flora. The machine learning model based on initial gut microbiota data can accurately predict the changes in Bifidobacterium, which sheds light on personalized nutrition intervention and precise modulation of the intestinal flora.
研究9天补充益生元对肠道微生物群结构和功能的影响，并基于初始肠道微生物群数据建立机器学习模型，以预测益生元摄入后双歧杆菌的变化。
采用随机双盲自身对照设计，35名健康志愿者被要求连续9天每天摄入16g低聚果糖（FOS）或低聚半乳糖（GOS）。通过16S rRNA基因高通量测序研究摄入益生元后肠道微生物群的变化。利用PICRUSt推断细菌群落功能模块之间的差异。基于初始肠道微生物群数据的随机森林模型用于识别益生元摄入5天后双歧杆菌的变化，然后建立一个连续指数来预测双歧杆菌的变化。使用GOS干预9天后收集的粪便样本数据对模型进行验证。
QIIME分析粪便样本显示，FOS干预5天降低了肠道菌群的α多样性，在第9天有所反弹；在GOS组中，干预期间肠道微生物群的α多样性逐渐降低。FOS和GOS补充剂均未引起肠道微生物群β多样性的显著变化。预测模型的曲线下面积（AUC）为89.6%。连续指数能够成功预测双歧杆菌的变化（R=0.45，P=0.01），且预测准确性通过验证模型得到验证（R=0.62，P=0.01）。
短期益生元干预可显著降低肠道菌群的α多样性。基于初始肠道微生物群数据的机器学习模型能够准确预测双歧杆菌的变化，这为个性化营养干预和肠道菌群的精准调节提供了思路。
