The tambaqui (Colossoma macropomum), also known as cachama or gamitana, is one of the most important Amazonian fish for aquaculture in South America and Asia. Improving its performance and well-being is key to the sector’s sustainability. Now, a pioneering study reveals an innovative way to achieve this: using artificial intelligence to “read” the fish’s stress through its skin color and genetically select the most resilient individuals.
The research, focused on the variation of skin color in tambaqui, not only confirms that the darkening of the cachama’s belly is an indicator of stress but also demonstrates that this trait is heritable and can be selected for without compromising growth.
Key conclusions
- The intensity of the “countershading” coloration on the tambaqui’s belly is a visible and reliable indicator of its stress response.
- A computer vision system (CVS) with artificial intelligence was developed to automate the measurement of this coloration for large-scale phenotyping with high accuracy (88.2%).
- The stress response, measured through coloration, is a trait with moderate to high heritability (h² = 0.40), making it an excellent criterion for genetic selection.
- It is possible to select fish for stress resistance (based on color) without negatively affecting their growth traits, such as final weight.
- The stress-linked hormone α-MSH causes an 80% expansion in melanophores, confirming the physiological basis of the color change.
Color as a mirror of stress in tambaqui
Fish, like other animals, change color for various reasons. This study, published by researchers from São Paulo State University (Unesp), the Brazilian Agricultural Research Corporation (EMBRAPA), and the University of Wisconsin, focused on two key mechanisms in the tambaqui:
- Morphological change (Long-Term): To observe this effect, researchers subjected a group of fish to a controlled stress condition, moving them from 200 m² earthen ponds to much smaller 1.8 m³ confinement tanks. Photographs taken before and after a 10-day period showed a clear result: the ventral area of the fish, characterized by a color pattern known as “countershading,” darkened significantly. Digital analysis confirmed a notable increase in the proportion of black pixels, demonstrating that prolonged stress visibly alters pigmentation.
- Physiological change (Short-Term): To understand the immediate response, the team investigated the effect of the alpha-melanocyte-stimulating hormone (α-MSH), known for its link to stress in fish. Upon exposing tambaqui scales to a solution with this hormone, they observed a rapid and drastic reaction under the microscope. In just 30 minutes, the percentage of expanded melanophores (pigment cells) increased by 80%, resulting in a darkening at the cellular level. This confirms the physiological basis connecting the hormonal stress response to an almost instantaneous color change.
“First, we observed that under stressful conditions—that is, in a more confined environment than normal—the fish became darker. We then found that adding a stress-linked hormone also changed the color of the scales. Next, we trained software with over 3,000 images to determine a stress threshold that could guide fish farmers and genetic selection programs, as we observed this is a hereditary characteristic,” explains Diogo Hashimoto, a professor at Unesp’s Aquaculture Center (Caunesp), who coordinated the study.
Automating assessment: Artificial Intelligence at the service of aquaculture
Manually measuring color or cortisol levels in thousands of fish is subjective, costly, and impractical for a genetic improvement program. This is where technology makes a difference. The researchers developed a computer vision system (CVS) to automate the process.
How does this system work?
The team used a deep learning approach. First, they took thousands of images and videos of two different tambaqui populations. Then, an expert manually segmented the area of interest (the countershading zone) in 4,500 of these images to “teach” a neural network (the DeepLab V3 model) to identify this region on its own.
The result was a robust system capable of analyzing an image and automatically delimiting the coloration zone with an 88.2% accuracy (measured by the IoU metric). This tool allows for the rapid and efficient collection of objective, large-scale phenotypic data, such as the percentage of black pixels (%BP) or the mean pixel intensity (MPI).
Stay Always Informed
Join our communities to instantly receive the most important news, reports, and analysis from the aquaculture industry.
Implications for cachama genetic improvement
With this powerful tool, the study took the most important step: estimating the genetic parameters of these coloration traits.
Heritability, which indicates what portion of a trait’s variation is due to genes, was found to be moderate to high for the color characteristics evaluated by the CVS. Values of h2=0.456 were estimated for the percentage of black pixels and h2=0.494 for the mean pixel intensity. In practical terms, this means that the stress response, manifested in color, is a trait that can be effectively improved through genetic selection.
What does this mean for fish farmers?
Perhaps the most relevant finding for the industry is the genetic correlation between coloration and growth. The study found that the genetic correlation between the percentage of black pixels and traits like final weight and weight gain was weak or practically null.
This breaks down a potential barrier to improvement: producers can select for genetically more stress-resilient tambaqui without fear of sacrificing their growth potential. This “precision breeding” approach not only enhances animal welfare but also opens the door to selecting fish for appearance traits, which are increasingly valued by consumers.
Conclusion
This study represents a significant advancement for tambaqui aquaculture and a brilliant example of how technology can modernize breeding programs. By establishing a clear genetic link between coloration and stress response, and by developing an artificial intelligence tool to measure it on a large scale, a new path is opened to produce more robust, healthy, and well-adapted fish for farming conditions, laying the groundwork for a more efficient and sustainable industry.
“The AI tool can be used to monitor the stress of farmed fish at a time when demands for animal welfare are growing. By simply evaluating photos of the animals, it would be possible to obtain this measurement and improve practices when necessary, such as reducing the stocking density in tanks,” the researcher concluded.
Contact
Diogo T. Hashimoto
São Paulo State University – Unesp, Aquaculture Center of Unesp
14884-900 Jaboticabal, SP, Brazil
Email: diogo.hashimoto@unesp.br
Reference
Lemos, C. G., Garcia, B. F., Filho, M. S., Arango, J. R., Butzge, A. J., Shiotsuki, L., Freitas, L. E. L., Rezende, F. P., Urbinati, E. C., Rosa, G. J., & Hashimoto, D. T. (2025). Deep learning approach for genetic selection of stress response in the Amazon fish Colossoma macropomum. Aquaculture, 609, 742848. https://doi.org/10.1016/j.aquaculture.2025.742848
Editor at the digital magazine AquaHoy. He holds a degree in Aquaculture Biology from the National University of Santa (UNS) and a Master’s degree in Science and Innovation Management from the Polytechnic University of Valencia, with postgraduate diplomas in Business Innovation and Innovation Management. He possesses extensive experience in the aquaculture and fisheries sector, having led the Fisheries Innovation Unit of the National Program for Innovation in Fisheries and Aquaculture (PNIPA). He has served as a senior consultant in technology watch, an innovation project formulator and advisor, and a lecturer at UNS. He is a member of the Peruvian College of Biologists and was recognized by the World Aquaculture Society (WAS) in 2016 for his contribution to aquaculture.
