A machine learning model using gut microbiome data for predicting changes of trimethylamine-N-oxide in healthy volunteers after choline consumption / 南方医科大学学报

<p><b>OBJECTIVE</b>To establish a machine learning model based on gut microbiota for predicting the level of trimethylamine N-oxide (TMAO) metabolism in vivo after choline intake to provide guidance of individualized precision diet and evidence for screening population at high risks of cardiovascular disease.</p><p><b>METHODS</b>We quantified plasma levels of TMAO in 18 healthy volunteers before and 8 h after a choline challenge (ingestion of two boiled eggs). The volunteers were divided into two groups with increased or decreased TMAO level following choline challenge. Fresh fecal samples were collected before taking fasting blood samples for amplifying 16S rRNA V4 tags, and the PCR products were sequenced using the platform of Illumina HiSeq 2000. The differences in gut microbiata between subjects with increased and decreased plasma TMAO were analyzed using QIIME. Based on the gut microbiota data and TMAO levels in the two groups, the prediction model was established using the machine learning random forest algorithm, and the validity of the model was tested using a verified dataset.</p><p><b>RESULTS</b>An obvious difference was found in beta diversity of the gut microbota between the subjects with increased and decreased plasma TMAO level following choline challenge. The area under the curve (AUC) of the model was 86.39% (95% CI: 72.7%-100%). Using the verified dataset, the model showed a much higher probability for correctly predicting TMAO variation following choline challenge.</p><p><b>CONCLUSION</b>The model is feasible and reliable for predicting the level of TMAO metabolism in vivo based on gut microbiota.</p>

Distribution characteristics of trimethylamine N-oxide and its association with gut microbiota / 南方医科大学学报

<p><b>OBJECTIVE</b>To analyze the distribution of trimethylamine N-oxide (TMAO) in healthy adults with different risk factors and explore its association with gut microbiota.</p><p><b>METHODS</b>We collected fasting blood samples and fresh fecal samples from 181 subjects without atherogenesis in the carotid arteries. Plasma TMAO levels of the subjects were determined using stable isotope dilution liquid chromatography-tandem mass spectrometry (LC-MS). The fecal DNA was extracted, and the 16S rRNA V4 tags were amplified and sequenced by Illumina HiSeq 2000. The association between TMAO and classical cardiovascular risk factors were analyzed. Gut microbial community structure was analyzed with QIIME, and LEfSe was used to identify the biomarkers.</p><p><b>RESULTS</b>The median (IQR) TMAO level was 2.66 (1.96-4.91) µmol/L in the subjects. TMAO level was significantly correlated with body mass index and operational taxonomic units (OTU). Individuals with high TMAO levels were found to have abundant Clostridiales, Phascolarctobacterium, Oscillibacter, and Alistipes but less abundant Anaerosprobacter.</p><p><b>CONCLUSION</b>Chinese subjects have in general low levels of TMAO. TMAO levels are not significantly correlated with the classical cardiovascular risk factors or the gut microbial structures.</p>

Methods, challenges and opportunities for big data analyses of microbiome / 南方医科大学学报

Microbiome is a novel research field related with a variety of chronic inflamatory diseases. Technically, there are two major approaches to analysis of microbiome: metataxonome by sequencing the 16S rRNA variable tags, and metagenome by shot-gun sequencing of the total microbial (mainly bacterial) genome mixture. The 16S rRNA sequencing analyses pipeline includes sequence quality control, diversity analyses, taxonomy and statistics; metagenome analyses further includes gene annotation and functional analyses. With the development of the sequencing techniques, the cost of sequencing will decrease, and big data analyses will become the central task. Data standardization, accumulation, modeling and disease prediction are crucial for future exploit of these data. Meanwhile, the information property in these data, and the functional verification with culture-dependent and culture-independent experiments remain the focus in future research. Studies of human microbiome will bring a better understanding of the relations between the human body and the microbiome, especially in the context of disease diagnosis and therapy, which promise rich research opportunities.