AI and Microbiome Analysis in Aquaculture: From Precision Science to Practical Tools

Microbiome research is a mature yet still rapidly evolving field. Today it is widely applied in aquaculture, animal welfare as well as in terrestrial livestock systems. The core idea is simple: by studying the microbial communities living in the gut of animals (or in the feces they produce) we gain insight into their health and the overall performance of the feed and environment we provide. This same logic sits behind a new generation of practical tools that aim to make microbiome monitoring easier to apply outside the lab, including approaches developed by companies such as HORAIZON Technology BV (www.horaizon.ai) and its aquaculture-facing solution - AquaGut (www.aquagut.com). In this post we describe HORAIZON’s approach and outline where rapid microbiome monitoring may offer practical value in RAS.

Until now, Next Generation Sequencing (NGS) techniques have been the gold standard for microbiome analysis. These methods allow researchers and practitioners to identify the dominant microbial taxa present in a sample and correlate that to something else happening in the animal or the system. By correlating dominant taxa with physiological or even environmental outcomes, we can infer how gut activity relates to nutrition and disease inside a closed system such as a recirculating aquaculture (RAS) farm.

Technological progress has also reduced turnaround times and cost. In commercial applications, microbiome community composition can now be analyzed within one to two days. While this is fast by laboratory standards, it can still be slow in highly dynamic RAS systems where things can change in hours.

Two Microbial Worlds in Recirculating Aquaculture Systems

In RAS, we are effectively managing two tightly coupled microbial ecosystems at the same time: the gut microbiome shaped by diet and health and the microbial communities living in the biofilters, mechanical filters, pipes, and tank surfaces and so on.

These two worlds are not independent. Research has shown that changes in fish gut microbiome composition can influence the microbial communities in filtration units and vice versa. The presence, absence or operating mode of key components (particularly mechanical filtration systems that remove organic matter and bacteria from the water) has a direct impact on both water quality and fish gut microbiomes.

The main challenge in applying microbiome science practically in RAS is response time. Many water quality and fish health issues develop rapidly, often within hours. These problems are frequently triggered by equipment failure or changes in operational performance. Once filtration efficiency drops then water quality can deteriorate quickly, leading to stress or disease outbreaks in fish showing up days later.

Traditional microbiome analytics may be too slow for operational decision-making especially during emergencies. However, several rapid assessment techniques already exist that can provide coarse indicators of bacterial activity. These methods do not tell us which bacteria are present or what they are doing at a functional level, but they do indicate whether bacteria are present and in what abundance.

When combined with water quality data and fish health observations, even low-resolution bacterial indicators can help farmers build an intuitive understanding of microbial dynamics without requiring detailed taxonomic knowledge.

A Middle Ground: AI-Based Pattern Recognition for Microbial Function

This is where a practical approach developed by the Dutch company HORAIZON Technology BV becomes interesting. In our view, it has the potential to fill the gap between slower, high-resolution sequencing and fast but low-information indicator methods.

HORAIZON’s workflow is intentionally simple. Fecal samples are smeared onto a printed template, the template is photographed and an artificial vision system analyzes the uploaded image and provides an assessment. In aquaculture, HORAIZON presents this capability through AquaGut.Rather than identifying specific taxa, the AI recognizes visual patterns in the smear and associates them with broader functional signals. In other domains, similar approaches are sometimes discussed in the context of “enterotypes” (community patterns linked to gut functions and dietary profiles). Enterotype classification was originally developed for gut environments and has been extensively studied in humans and terrestrial animals. It can be used to distinguish whether a gut ecosystem is driven mainly by protein- and lipid-rich diets, carbohydrate-rich diets, or mixed nutritional inputs.

Recirculating systems add an extra layer of complexity, because samples from drum filters, biofilters or pipelines reflect a mixture of fish-derived inputs and microbial processes within the system itself. That complexity is exactly why pattern recognition combined with AI could be useful as a rapid and affordable way to monitor functional microbial states in RAS - especially to detect shifts and trends over time.

Potential applications include identifying a “nitrifying-like” functional group and tracking its performance across system components, detecting sulfur-associated functional communities in specific pipelines or filters, exploring correlations between fish gut patterns and system-level microbial signals and studying whether shifts in system microbiology precede or coincide with pathogen outbreaks.

Below is a simplified comparison of the main microbiome-related techniques discussed:

Conclusion

In recirculating aquaculture systems, the main value of HORAIZON approach is speed and repeatability. The opportunity is to use it as a monitoring layer (fast enough for operational tempo) while continuing to rely on sequencing and established methods when deeper resolution is needed. In that sense, it does not replace existing analytics. It expands the toolbox and makes time-series snapshots of microbial dynamics more practical in RAS.

If you are interested in this technology, you can contact me or HORAIZON directly:

www.horaizon.ai www.aquagut.com