ISSUE FOCUS FEED & ADDITIVE MAGAZINE February 2026 59 EARLY DETECTION OF HEALTH PROBLEMS In 2019, a trial was conducted looking at the response time of certain blood biomarkers to a coccidiosis challenge. Potassium and carotenoids began to shift 3-7 days before any other obvious or macroscopic symptoms becoming apparent. This rapid response sparked the idea for an early warning system for coccidiosis. The hypothesis was proposed that with enough data, machine learning could be used to create a classifier model with a forecasting capacity for coccidiosis. MACHINE LEARNING Verax™ uses supervised machine learning to create classifier and regressor models. There are currently many tens of thousands of data points in the database, gathered from commercial broilers with a naturally occurring prevalence of coccidiosis. To create the model, the data set was split into two sections; 60% used for training, and 40% used for validation. All the birds with coccidiosis were identified and a biomarker profile was created which predicted that phenotype. The model was then validated on the other subset of birds. Over time and with more data, especially from birds that have coccidiosis, the accuracy of the model increases and permits the identification of specific Eimeira species. This principle was applied in practice on a farm in the US. Blood samples were taken from birds on four different farms on day 14. The blood analysis results were used to predict that two of the farms would have a coccidiosis outbreak later, and the other two would not. A second visit to the farms on day 28 confirmed the predictions. Although the model is not 100% accurate yet, there is a very strong association with excellent statistical performance in terms of false positive and false negative rates on the forecasting ability of the model. Figure 8 shows an example of the user interface in Verax™ for tracking flocks, including coccidiosis scores, over time. CONCLUSIONS • The importance of data cannot be overstated. Data science will continue to unlock new opportunities for poultry producers if a more systematic approach is taken towards data handling, capturing, and processing. • New technologies and tools are allowing nutrition to be more precise than ever before. Nutritional optimization is getting easier with improved monitoring and shorter feedback loops. • By collecting and analyzing data from a variety of sources, nutritionists are better able to unlock new levels of bird performance. New tools like large language models are making it much easier to ingest unstructured data sources, but the data must be accessible to begin with. • Blood biomarkers can be used to predict disease outbreaks earlier than ever before. Biomarker Impact Biomarker Impact History of Results History of Results Increases score: Increases score: Coccidiosis-GAC score Coccidiosis-GAC score Moderate Moderate ↑ +9 from sampling september 23, 2021 ↑ +9 from sampling september 23, 2021 Our GAC score indicates your birds are experiencing moderate effects by GAC on biomarkers. Biomarkers that are predicted to be affected by GAC (based on the analysis of this sampling) are displayed Our GAC score indicates your birds are experiencing moderate effects by GAC on biomarkers. Biomarkers that are predicted to be affected by GAC (based on the analysis of this sampling) are displayed 100 100 0 90 90 0 Decreases score: Figure 8. Verax™ user interface
RkJQdWJsaXNoZXIy MTUxNjkxNQ==