Deriving novel traits based on data from sensors and other technologies

Authors: KOLTES, JAMES - Iowa State University; JAMES, LEO - Iowa State University; MAYES, MARY - Iowa State University; COOPER, CORI - Iowa State University; PARKER-GADDIS, KRISTEN - Council On Dairy Cattle Breeding; Vanraden, Paul; Baldwin, Ransom - Randy; SANTOS, JOSE - University Of Florida; TEMPELMAN, ROBERT - Michigan State University; WHITE, HEATHER - University Of Wisconsin; PENAGARICANO, FRANCISCO - University Of Wisconsin; WEIGEL, KENT - University Of Wisconsin; VANDEHAAR, MICHAEL - Michigan State University

Submitted to: Journal of Dairy Science
Publication Type: Abstract Only
Publication Acceptance Date: 3/6/2024
Publication Date: 6/25/2024
Citation: Koltes, J.E., James, L., Mayes, M., Cooper, C., Parker-Gaddis, K.L., Van Raden, P.M., Baldwin, R.L., Santos, J.E., Tempelman, R.J., White, H.M., Penagaricano, F., Weigel, K., Vandehaar, M.J. 2024. Deriving novel traits based on data from sensors and other technologies [abstract]. Journal of Dairy Science. 107(Suppl. 1):39(abstr. 1191).

Interpretive Summary:

Technical Abstract: Feed efficiency is an economically and environmentally important trait in dairy cattle. Estimating feed efficiency requires individual animal feed intake, milk yield and components, and ideally information about body weight and body condition score. Unfortunately, apart from milk traits, these data are rarely available on commercial farms as they are cost prohibitive. Thus, there is a need to identify new indicator traits for feed efficiency. Milk testing and sensor data represent existing information streams on dairy farms that may be able to serve as proxies for traits composing feed efficiency. Multiple experiments were conducted to evaluate the utility of these data. Five different types of sensors measuring activity, rumen traits, and temperature were evaluated for their phenotypic association with feed efficiency traits. Heritabilities were estimated for three sensor-based traits from repeated records for eartag and milking collar-based sensors, respectively. Estimated heritabilities were 0.21 (eartag activity), 0.14 (eartag rumination), and 0.17 (collar activity). Milk metabolites were also investigated to identify metabolites associated with or predictive of feed intake. Extreme cows for residual feed intake (RFI; top and bottom 15%) were selected. Three assays were performed: gas chromatography mass spectrometry, liquid chromatography mass spectrometry, and fatty acids inferred from Fourier Transformed Infrared Spectrometry. A total of 33 metabolites and 10 pathways were identified as statistically associated with feed intake. Milk fatty data were predictive of feed intake with an R^2 of 0.86 and a concordance correlation coefficient of 0.92 based on an across cow cross-validation approach using a random 25% of cows, independent of the training set, as the testing set. Although further validation is needed, these studies indicate that sensor data could be useful as indicator traits for feed intake and milk-based data may be predictive of feed intake.