Symposium review: Challenges and opportunities for evaluating and using the genetic potential of dairy cattle in the new era of sensor data from automation

摘要

Sensor data from automation are becoming availableon an increasingly large scale, and associated researchis slowly starting to appear. This new era of sensor datafrom automation leads to many challenges but also newopportunities for assessing and maximizing the geneticpotential of dairy cattle. The first challenge is data quality,because all uses of sensor data require careful dataquality validation, potentially using external references.The second issue is data accessibility. Indeed, sensordata generated from automation are often designed tobe available on-farm in a given system. However, tomake these data useful-for genetic improvement forexample-the data must also be made available offfarm.By nature, sensor data often are very complexand diverse; therefore, a data consolidation and integrationlayer is required. Moreover, the traits we wantto select have to be defined precisely when generatedfrom these raw data. This approach is obviously alsobeneficial to limit the challenge of extremely high datavolumes generated by sensors. An additional challengeis that sensors will always be deployed in a context ofherd management; therefore, any efforts to make themuseful should focus on both breeding and management.However, this challenge also leads to opportunities touse genomic predictions based on these novel data forbreeding and management. Access to relevant phenotypesis crucial for every genomic evaluation system.The automatic generation of training data, on both thephenotypic and genomic levels, is a major opportunityto access novel, precise, continuously updated, andrelevant data. If the challenges of bidirectional datatransfer between farms and external databases can besolved, new opportunities for continuous genomic evaluationsintegrating genotypes and the most current localphenotypes can be expected to appear. Novel conceptssuch as federated learning may help to limit exchangeof raw data and, therefore, data ownership issues,which is another important element limiting access tosensor data. Accurate genome-guided decision-makingand genome-guided management of dairy cattle shouldbe the ultimate way to add value to sensor data fromautomation. This could also be the major driving forceto improve the cost-benefit relationship for sensorbasedtechnologies, which is currently one of the majorobstacles for large-scale use of available technologies.