Zebrafish eggs undergo typical developmental stages, yet even within the same clutch, embryos progress at varying rates. Utilizing artificial intelligence, researchers can analyze disparities in terms of developmental pace, standard phases, and structural discrepancies. Credit: Nikan Toulany, Patrick Müller
The development of pet embryos from fertilized cells to functional organisms involves distinct stages. This intrinsic process is predominantly governed by physical factors and follows a common pattern observed across diverse animal species.
Nevertheless, variations exist in the specifics, both within eggs of the same species and across different species. For example, the timing of completion for various larval stages may vary. These deviations in embryonic development are believed to play a pivotal role in evolution by giving rise to new traits, thereby fostering biodiversity and adaptive changes.
Therefore, studying animal fetal growth is essential for gaining a deeper understanding of biological mechanisms. The question arises: How can variations in fetal development, such as the timing of evolutionary stages, be accurately and objectively documented? Researchers at the University of Konstanz, led by systems biologist Patrick Müller, are developing and employing artificial intelligence (AI) techniques for this purpose.
In their recent publication in Nature Methods, they introduce an innovative method that can identify distinct stages across species boundaries and promptly assess the speed of developmental processes.
Each zygote possesses unique characteristics.
The foundational knowledge of dog embryogenesis and developmental stages stems from research involving meticulous microscopic observation and description of embryos from various stages. This painstaking manual work has enabled many avian species to have access to reference materials featuring idealized representations of different fetal stages.
However, embryos may not always exhibit exact correspondence between their microscopic appearance and schematic illustrations. According to Müller, the transitions between different stages are more gradual than abrupt, making manual delineation of developmental stages challenging and somewhat subjective even for experts.
The variability in the timeline of fetal development further complicates matters. Müller notes, “Various factors, such as temperature, can influence the timing of fetal development.” The AI-assisted approach developed by him and his team marks a significant leap forward.
By training their Twin Network with over 3 million images of healthy fish embryos, the researchers were able to swiftly determine the developmental age of diverse zebrafish embryos using the resultant IoT model.
Precision, Significance, and Purpose
The researchers successfully demonstrated that AI can autonomously identify crucial milestones in zebrafish embryogenesis and individual developmental stages without human intervention. In their study, the researchers leveraged the AI system to investigate the temperature sensitivity of zebrafish fetal development and assess the developmental stages of embryos.
While the AI was primarily trained on images of normally developing embryos, it also exhibited the capability to detect abnormalities, which may arise spontaneously in a small fraction of cases or due to environmental toxins.
Expanding their approach to other pet species such as sticklebacks and the evolutionary distinct insect Caenorhabditis elegans, the researchers highlighted the versatility of their Twin Network-based method in assessing fetal development across various animal species when adequate visual data are available. Müller elucidates, “Our system functions with a clear objective, even in the absence of analytical data for the pet species, showcasing its potential for unraveling the evolutionary trajectories and developmental patterns of previously uncharacterized avian species.”
