Metabolic modeling for aquaculture
Doctoral thesis
Published version
Date
2024Metadata
Show full item recordCollections
- Doctoral theses (KBM) [136]
Abstract
Farmed Atlantic salmon is Norway’s second lagest export, and by 2030 aquaculture will provide most of the world’s food fish. Salmon is a highly resource-efficient production animal, but scarce feed resources require action to adapt new, sustain able diets. Over the last few decades, feed formulations have shifted from being mostly based on fish to using plant sources, which has brought new challenges, and
recipes have become increasingly complicated. However, in the volatile market for raw materials, the most cost-effective feed composition is a moving target. This applies both from an economic and sustainability perspective.
Meat production is basically a kind of chemical processing, where the sophisticated chemical machinery in the animals’ metabolism break down raw materials and build up new products. These chemical processes are carried out by molecular machines called which are coded for in the genome. With knowledge of which enzymes an organism’s genes code for, the metabolism’s chemical repertoire can also be mapped. Methods for systematising such a network of chemical reactions have been developed and used both in medicine and bioprocess technology, but until now little has been used in animal production and other production biology.
This thesis applies systems biology methods to study the salmon’s metabolism.
In the first paper, we develop SALARECON, a constraint-based metabolic model for salmon, and the first genome-scale metabolic reconstruction for a production animal. We use this to study the relationship between growth rate and oxygen saturation in the water, as well as to predict growth-limiting amino acids for growth on different protein sources.
The second paper focuses on functional interpretations of gene expression data and using gene expression data to remove low-expression reactions to generate sample-specific models with bounded solution spaces. Starting from SALARECON, six methods were used to create mathematically and stoichiometrically consistent submodels for each of 208 samples from a feed-switch experiment. To compare the models functionally, we used a test framework for chemical processes,
where SALARECON was able to perform 121. The relevance of genes for certain processes was determined through simulations with SALARECON, and this was combined with the gene expression data to form a reference basis for each of the samples. Each sub-model was characterized functionally, and compared with the reference of the corresponding sample. Of the six methods, GIMME, iMAT and
INIT were best able to produce models according to the criteria. The study shows that sample-specific sub-models can provide more precise functional insight than a general model, and that these methods developed for humans are transferable to other non-mammals.
In the third paper, we extended SALARECON with flexible fat composition to
the model SimSaLipiM and used this for simulated feeding experiments. The model was extended with synthesis and breakdown of fatty acids, as well as flexible fatty acid composition in the biomass to reflect the observed variation. In order to adapt the model to feed composition data, the model can be regenerated based on a list of fatty acids. We formulated feed intake in the model, based
on chemical compositions of feed ingredients, and used CO2 footprint of ingredients as weights to identify sustainable feed recipes. Optimal feed compositions were highly context-dependent and location, availability of ingredients and how the criterion for optimality has a major impact on the result.
In summary, this thesis points the way to applications of metabolic modeling in research on salmon. Mathematical models serve as lenses through which we interpret data and explore the consequences of underlying assumptions. They allow us to analyze complex systems, evaluate alternative strategies, and identify causal effects, bridging theory and empirical observations, as shown in the development
of feed recipes and gene expression data, integrating modeling in future experiments on salmon will both contribute to enriching the results and to developing the model framework further. Laksenæringen er Norges nest største eksportnæring, og innen 2030 vil over halvparten av verdens matfiskforbruk komme fra oppdrett. Atlanterhavslaks er et svært ressurseffektivt produksjonsdyr, men begrensede fôrressurser krever handling for å tilpasse nye, bærekraftige dietter. Oppskriftene har blitt mer og mer kompliserte, og med et råvaremarked i stadig omveltning, varierer det hva som er den mest lønnsomme fôrsammensetningen, både økonomisk og i bærekraftperspektiv.
Kjøttproduksjon er i bunn og grunn en slags kjemisk prosessering, hvor det sofistikerte maskineriet i dyrenes stoffskifte bryter ned råstoff og bygger opp nye produkter. Enzymer, de molekylære maskinene som utfører arbeidet, er kodet for i genene. Med kunnskap om hvilke enzymer en organismes gener koder for, kan også stoffskiftets kjemiske repertoar kartlegges. Metoder for å systematisere et slikt nettverk av kjemiske reaksjoner er utviklet og anvendt både i medisin og bioprosessteknologi, men inntil nå har det vært så godt som fraværende i dyreproduksjon og annen produksjonsbiologi.