Predicting Conception Rate In Cow Based On The Dynamics Of Cervical And Fecal Microbial Community

With the rapid development of economy,beef consumption has increased rapidly.in China;however,the beef cattle industry falls far behind compared to the market demand.Cattle are monotonous animals with a long breeding cycle,which means low reproductive capacity,which is one of the most important bottlenecks of the beef cattle industry.Microbiome plays a vital role in heifer's genital system.Severe pathogenic infection leads to reproductive disorder and low feralization rate.In order to explore the microbiome dynamics during the whole pregnancy process and investigate the host-microbe interactions,we launched a study on the bacterial and fungal community in both cervix and feces using 16 s rRNA and ITS sequencing methods.Meanwhile,we established models with pre-breeding microbiome community data to predict the fertilization outcome of heifers.In this study,the main contents are as follows:1.A total of 20 heifers' pre-breeding cervical swabs and fecal samples were selected and analyzed to explore the fungal communities in this study.The ITS hypervariable region was targeted and sequenced using Illumina Miseq platform.The fertilization outcome(Open and Bred groups)didn't have an impact(P > 0.05)on alpha or beta diversity of cervical and fecal fungal community.Random forest-based fertilization outcome predictive models show a higher accuracy rate achieved from feces than that from cervical fungi(AUC: 0.86 vs.0.65).2.Fecal samples were collected from a total of 68(56 Bred and 12 Open)heifers on pre-breeding.Amplicons targeting 16 s rRNA V4 region were sequenced by Illumina Miseq platform.Based on the diversity(alpha and beta)analysis,alpha and beta diversities showing no significant differences(P > 0.05)between Bred and Open groups at prebreeding stage.With the understanding of pre-breeding bacteria in both Bred and Open heifers,we constructed a random forest model,the predictive model achieved an accuracy of 99.2%(AUC = 0.992)on the fertilization outcome.An additional 16 s rRNA V4 sequencing dataset containing thirty-one samples was used as validation dataset.Comparing community diversity between the two datasets,alpha diversity didn't show significant difference,but beta diversity in the later was significantly different(P < 0.001)from this study.The predictive model was established by Random forest based on the data from both studies,with an accuracy of 90.5%(AUC=0.905).3.The published dataset published by our group was download from database.The dataset containing 16 s rRNA V4 sequences of bovine cervical microbiota at pre-breeding(n=72),first-trimester(n=72),second-trimester(n=72)and third-trimester(n=56).Using bioinformatic methods,cervical bacterial dynamics during the whole breeding cycle were analyzed.With the understanding of pre-breeding bacterial structures in both fertilized(Bred)and unfertilized heifers(Open),we constructed a fertilization outcome predictor using random forest model.Alpha and beta diversities showed no significant differences(P > 0.05,Shannon index and the number of observed OTU's)between Open and Bred groups at all four-time points.At the community level,significant differences in alpha diversity indices of the cervical microbiome were observed over time.The community richness(number of observed features)and diversity(Shannon Index)experienced an ‘up and down' trend.Using machine learning algorithm,the predictive model achieved an accuracy of 84.9%(AUC = 0.849)on the fertilization outcome.4.Fecal samples were collected from a total of 20(10 Open and 10 Bred)heifers on pre-breeding and shotgun metagenomic sequencing was performed.Using bioinformatic methods,the difference of taxonomy composition and functional genes of bovine feces between Bred and Open.Bacteria and archaea are dominant bacterial community of bovine feces at pre-breeding stage.The Lef Se analysis shows that Rhodococcus equi is a potential pathogen caused failure of cow breeding.The predictive model was established by Random forest based on the gene cluster constructed from metagenomic data,with an accuracy of 84%.5.A total of six standardized genomic annotation descriptive index,including gene density,distribution of gene length,transcript number,distribution of transcript length,distribution of exon number,and statistics of annotated UTR of each transcript,were selected to design and develop a new software for genomic annotation information visualization.The GFFview was developed with Python language and carry six standardized metagenomic annotation descriptive Index.The GFFview software is now freely accessible to the public on Cloud server.Our study described the heifer's longitudinal microbiome dynamics in cervix during the pregnancy cycle according to 16 s rRNA gene target sequencing method,which helps for understanding the relationship between heifer' cervical and fecal bacteria and reproduction of heifers.In addition,we established five random forest models and the comparative result indicated that the model based on fecal bacteria shows the highest prediction accuracy.This result has great practical significance for the management of reproduction of heifer.