Advanced statistical modeling of complex traits related to subacute ruminal acidosis in dairy cows
von André Mensching
Datum der mündl. Prüfung:2020-07-14
Erschienen:2021-05-17
Betreuer:Dr. Ahmad Reza Sharifi
Gutachter:Dr. Ahmad Reza Sharifi
Gutachter:Prof. Dr. Jürgen Hummel
Gutachter:Prof. Dr. Dr. Sven Dänicke
Gutachter:Prof. Dr. Nicolas Gengler
Dateien
Name:2021_Mensching_Dissertation_eDiss.pdf
Size:8.31Mb
Format:PDF
Description:Dissertation
Zusammenfassung
Englisch
Adequate feeding of lactating cows is particularly challenging for high-performing animals. In this respect, subacute ruminal acidosis (SARA) is considered the most important nutritional disease. Its emergence is promoted by an imbalanced design of the diet, where both an excess of easily fermentable carbohydrates and a deficiency in the physical structure of the feed can be critical. The pH parameters derived from continuous ruminal pH measurements are considered the gold standard for the diagnosis of SARA, but they cannot be measured on a large scale in agricultural practice. Since SARA corresponds to a subacute stage, no specific clinical signs can be detected at the animal level. However, associations with clinical signs at herd level are described, such as reduced feed intake, lower milk production efficiency and a higher risk of loose stools and claw diseases. Consequently, there is a need to identify indica-tors that allow for a comprehensive and precise monitoring of SARA to minimize economic losses and in particular to ensure animal welfare. The objective of this dissertation was therefore to investigate the associations between data that were measured in vivo in the reticulorumen, properties of the diet, various behavioral and blood parameters and the composition of the milk using different statistical methods, in order to contribute to the development of better indicators for SARA. First, the associations between ruminal pH parameters, feed properties and particularly the main milk components were investigated in a meta-analysis. To create a data basis, the results of 32 studies with continuous ruminal pH measurements in dairy cows were gathered. The main part of the analysis consisted of a systematic examination of potential predictors for ru-minal pH parameters using mixed multi-level meta-regression models. Significant associations between ruminal pH parameters and the protein and fat content of milk as well as the milk fat to milk protein ratio were confirmed. However, the associations can only be classified as a trend, since most of the observed variability of pH parameters is due to a high level of heterogeneity both within and between the individual studies which means that only a small proportion of the variance was explained by the predictor variables. In a further analysis, the ruminal pH development both in the reticulum and in the ventral rumen in the course of the day was examined based on data collected in an experimental sta-tion. The aim was to model the pH development using sensor-based records of the feed intake, water intake and rumination behavior. An extensive data preparation was crucial, whereby all available data were transformed into a uniform 1-minute resolution. In addition, signal transformation methods were used to model feed and water intake events over time. For the statistical evaluation, all data were analyzed in the form of high-resolution time series using linear mixed regression models. It was shown that the daily pH development is highly associated with the animal's individual feed intake and rumination behavior in the course of the day. Based on the previously obtained knowledge, a procedure was developed to predict the risk of suffering from SARA using milk mid-infrared (MIR) spectral data. The data were collected on 10 commercial farms and a total of 100 cows. This included reticular pH measurements, behavioral data, feed analysis data, performance and blood data. Furthermore, MIR spectral data as well as information on the main components of the milk and the fatty acid composition of the milk fat were available. Since sensitive sensor systems were used, which are prone to technically caused errors under the environmental conditions occurring in agricultural practice, a multivariate plausibility check was first developed to prepare the data set for downstream analyses. The aim was to develop a procedure to classify the individual observations of several simultaneously recorded sensor, blood and milk data into ‘physiologically normal’, ‘physiologically extreme’ and ‘implausible’. In the final analysis, the association between potentially SARA-indicating traits and the composition of the milk was examined using the processed data set. First, an innovative SARA phenotype (‘SARA risk score’, SRS) in the form of an index trait was developed. This SRS is based on information from intra-ruminal pH and temperature measurements, rumination and feed intake behavior as well as the milk performance. Using partial least squares regression models, a MIR-based prediction model with a moderate prediction quality could be established for the SRS. In addition, significant associations between the fatty acid profile of milk and the SRS were determined. On the one hand, well-known relationships between ruminal fermentation, animal behavior and milk constituents were verified using innovative statistical methods. On the other hand, it was shown that the MIR spectral data of milk, which are already routinely collected, offer an enormous potential for the characterization of the health status of lactating cows. The ob-tained findings in this work thus provide the basis for the development of a routine and com-prehensive SARA monitoring which can be applied in agricultural practice.
Keywords: Statistical modeling; Subacute ruminal acidosis; Dairy; Animal welfare