Study On Predicting Model Of Methane Production In Vitro By Ruminal Dominant Bacteria And Feed Components In Sheep

Methane is one of the important factors that produce greenhouse effect.Accurately predicting the production of methane from ruminant rumen is the basis for evaluating the greenhouse effect.In this study,high-throughput sequencing and qPCR analysis were used to analyze the effects of different diets on the dynamic changes of rumen microorganism in order to carry out the marker microorganism.The open-type fermentation system in vitro was used to determine the methane yield,finally,to establish an artificial neural network model for predicting the production of methane in vitro.Six healthy rumen fistula sheep（Xinjiang fine wool sheep×dorper sheep hybrid）were selected.Six kinds of diets with coarse straw（CS）and Leymus chinensis（LC）as forage,which were compounded with concentrate by the ratio of 1:9,3:7 and 5:5,were used to feed the sheep by the 6×6 latin square design.After 10 days of pre-feeding,rumen liquid was collected and high-throughput sequencing of rumen liquid bacteria was carried out.The results showed that the rumen microbes had changed significantly by feeding the different forage and different ratio diets.Compared to the effect of different forage,the effect of different ratio on rumen flora was more significant.The results showed that Papillibacter（PAP）,Ruminococcaceae_UCG-011,Ruminococcus₁,Bacteroidales_S24-7,Prevotella（PRE1）₁,Prevotellaceae_UCG-001、Rikenellaceae_RC9 and Christensenellaceae_R-7 were differentiated as the significantly different dominant bacteria by different diets.The six kinds of diets with different coarse roughness and different rough feed sources were tested in a certain order to simulate the dynamic changes of the rumen.During the change,the CH₄ yield of 60 kinds of mixed feed was measured by the open in vitro fermentation system,meanwhile quantifying the relative concentration of significantly different dominant bacteria.Then the database for modeling and testing was Successfully built.Under the condition of rumen dynamic changes by the six diets,there were a significantly multiple linear relationship between the CH₄ yield（mmol）in vitro and the feed nutrients of CP,CHO,NSC（g）and the rumen dominant bacteria of PRE1,PAP（%）:CH₄=（0.469±0.049）PRE1+（-4.122±0.73）PAP+（5.316±0.641）CP+（3.169±0.557）CHO+（-0.42±0.078）NSC+（-287.371±50.466）,R²=0.822,n=50,P<0.0001.On the other hand,the CH₄ yield in vitro was used as output variable and the feed nutrients of CP,CHO,NSC（g）and the rumen dominant bacteria of PRE1,PAP（%）were used as input variables to establish the BP neural network model.The best structure was 5-9-1 and it was more accurate than the multiple linear regression model.In this study,under the condition of rumen dynamic changes,predicted models with good accuracy were established by the relative concentration of rumen dominant bacteria and feed nutrients,which could provide some references for the future application of CH₄ prediction model.