Towards The Deciphering Of The Association Between Insomnia Disorder And Gut Microbiota Using Machine Learning Approach

ObjectiveTo investigate the association between insomnia disorder and gut microbiota and evaluate potential application of utilizing the gut microbiota as the non-invasive biomarkers for insomnia disorder diagnosis,we take advantage of the high throughput sequencing,novel bioinformatic pipeline and machine learning technology to pinpoint the involvement of gut flora with insomnia.MethodUpon approval by ethics committee of Jinan University,20 volunteers were enrolled at the first affiliated hospital of Jinan University(Guangzhou overseas Chinese hospital)in 2018.All of organic diseases and drug intake were excluded.According to Diagnostic Interview for Sleep Patterns and Disorder(DISP),20 volunteers were separated into two groups(Insomnia group and Normal Control group),and received polysomnography treatment.Fecal samples were collected by sterilized instruments followed with the DNA isolation using the Mobio PowerSoil? DNA Extraction kit according to the manufacture's instruction.With specific primer-set(338F-806R)targeting to the V3-V4 region of bacterial 16 S rRNA gene,all samples were sequenced on an Illumina HiSeq 2500 platform.The down-stream analysis was performed on R(3.5.1 version),Ubuntu 16.04.4 LTS,Python 2.7.14,Python 3.6.1 platform.Result1.Raw sequencing data were quality-controlled by Sequencing depth test,Species accumulation curve,and Good's Coverage,which are sufficient for the down-stream analysis.2.Significant microbiota structural difference between insomnia and healthy population was distinguished.3.BugBase and PICRUSt analysis indicated a significant functional difference of gut microbiota between insomnia and healthy population.4.The statistical analysis through spearman correlation,species plotted for co-occurrence network separately for each group suggested that the microbial community among two groups were significantly different.By applying the greedy clustering method,insomnia group was sub-divided into 4 sub-communities,while normal group had 5 sub-communities5.Both RDA and ANOISM analysis strongly determined that insomnia as a key factor plays an essential role in separation and clustering of the gut microbiota in each group.6.By applying the five-fold cross-validation on a random forest model,two optimal species markers selected with consideration of the lowest mean error rate and standard deviation were selected to identify insomnia patients successfully.7.Based on gut microbiota data,Artificial Neural Network could predict the levels of insomnia patients accurately.Conclusion1.Significant difference of gut microbiota on structure,function and network was explored between insomnia and healthy population.2.Two optional species as potential biomarker were selected to distinguish insomnia disorder successfully.3.Artificial Neural Network prediction model could estimate levels of insomnia and healthy population.