Abstract
In ovo sexing involves identifying chicken embryo sex before or during incubation to avoid euthanizing male chicks after hatching, enhancing animal welfare in the laying hen industry. Recently, researchers demonstrated the potential for non-invasive and early in ovo sexing through the analysis of volatile organic compounds (VOCs) emitted by eggs. However, a knowledge gap persists in comprehending the robustness of prediction models, the efficacy of faster acquisition techniques, and the day-to-day performance variations. In our study, we performed two experiments to fill these gaps. In Experiment 1, passive VOC extractions were performed on 110 eggs on incubation day 10 using sampling bags employing headspace sorptive extraction-gas chromatography-mass spectrometry (HSSE-GC-MS), proton transfer reaction-time-of-flight-mass spectrometry (PTR-TOF-MS), and selected ion flow tube-mass spectrometry (SIFT-MS). Prediction models were built using partial least squares-discriminant analysis (PLS-DA) and variable selection methods. As a result, prediction accuracies ranged from 57.6 % to 61.4 %, indicating no significant difference between the devices and highlighting the need for further optimizations. In Experiment 2, passive VOC samplings were performed on 42 eggs in glass jars during the initial 12 days of incubation using HSSE-GC-MS. Consequently, the optimized setup yielded higher accuracies ranging from 63.1 % to 71.4 %, revealing VOCs consistently elevated in relative abundance for a specific sex, and overall VOC abundance was higher in male embryos. Suggestions for future experiments to increase the accuracy of VOC in ovo sexing include active sampling with inert materials, expanding sample sets, and targeting consistent compounds.
Supplementary materials
Title
Tables A1 and A2_HSSE-GC-MS compounds
Description
Detailed overview tables of the identified VOCs in Experiments 1 & 2 in hatching eggs during incubation using HSSE-GC-MS.
Actions
Title
Tables B1 and B2
Description
PLS-DA prediction models obtained in Experiments 1 & 2 using different variable selection methods.
Actions