Round fish farm cages floating on a calm sea loch with hills behind
Photo: Alpin Stewart / Wikimedia Commons, CC BY-SA 2.0

vizmaya desk · 6 November 2026

Fish farm feed efficiency: what breeding actually delivered

50 to 70%

That is the share of a fish farm's costs that goes on feed.

Can you breed a fish that needs less of it?

Source: Abdullahi et al. 2026, Biology 15(20):1849, p1

1 · The bill

Feed is the biggest bill on a fish farm.

Worldwide, feed makes up 50 to 70% of production costs.

In Nigeria it is 60 to 70%. For mandarin fish in China it is over 50%.

Feed has a second cost. Where does it show up?

Source: paper p1, p2 (refs 4, 5)

2 · The water

Less feed per fish means less nitrogen and phosphorus in the water.

Too much of both can choke a water body. The word for it is eutrophication.

Cutting feed per kilo of fish helps the farm's wallet and the water at once.

How do you score a fish for this?

Source: paper p1, p2

Fish farm cages in the Sound of Raasay, seen from the shore
Photo: David Dixon / Wikimedia Commons, CC BY-SA 2.0

3 · The score

The score is simple: feed in, fish out.

Feed conversion ratio, or FCR, is the feed a fish eats divided by the weight it gains.

Lower is better. An FCR of 2.0 means 2 kg of feed for 1 kg of fish.

Example numbers to show the maths, not data from the paper.

Is a good score luck, or does it run in families?

Source: paper p2 (definition)

A heap of small round brown fish feed pellets
Photo: Narek75 / Wikimedia Commons, CC BY-SA 4.0

4 · The old way

For years, the fix was in the feed bag.

Better feed and better farm routines improved feed efficiency over past decades.

Genes add a different kind of gain. The review calls it a permanent, cumulative route to sustainability.

Does the trait really pass from parent to young?

Source: paper p2 (ref 10)

5 · The turn

Yes. Feed efficiency runs in families.

Scientists score this from 0 to 1. At 0, none of the difference between fish is passed on. At 1, all of it is.

Every species in the review's table scored above zero. The spread is wide: 0.08 in salmon, 0.743 in shrimp.

† The paper flags a weakness in this number: shrimp could not separate shared-environment effects, turbot gave no error range.

Why is the spread so wide?

Source: paper p3 (Table 1), p7

6 · The crack

Change the water or the diet, and the winners change.

In large yellow croaker, gene spots tied to FCR numbered 6 at 15 °C, none at 19 °C and 7 at 24 °C.

In shrimp, how families rank on a plant diet versus a fishmeal diet matches only partly: 0.646, where 1.0 is a perfect match.

Does the fastest grower also eat the least?

Source: paper p7 (ref 15), p8 (ref 32)

7 · The crack

Fast growers eat less in tilapia. In shrimp, the link is gone.

In tilapia, faster growth goes with a better FCR (-0.69). In shrimp, the link is about zero (-0.018).

RFI: feed eaten beyond what a fish's size and growth predict. † No error range reported.

A Nile tilapia held up, grey-silver with dark vertical bars on its tail
Photo: Raver Duane, U.S. Fish and Wildlife Service / Wikimedia Commons, public domain

So does breeding for it work?

Source: paper p7 (ref 19), p8 (refs 12, 29, 30)

8 · The result

Breeding works. Ten generations cut trout feed use by 17 to 20%.

Rainbow trout were bred for ten generations, for several traits at once. Feed per kilo of fish gained fell by 17 to 20%.

Tilapia from the GIANT strain needed 15.7% less feed per kilo than commercial strains on farms.

A rainbow trout with a pink stripe and dark spots, held at the water's edge
Photo: Chris Light / Wikimedia Commons, CC BY-SA 4.0

Can reading DNA make this faster?

Source: paper p11 (refs 55, 56, 58)

9 · The tool

Reading a fish's DNA picks better parents.

Genomic selection reads DNA markers, so you need not measure every fish's eating. In tilapia, it raised prediction accuracy by up to 34% over family records alone.

In sea bass, the reliability of breeding-value estimates rose by about 10% to 125%, depending on the trait.

It still needs a trained group of fish that were measured and genotyped (p11).

And editing genes directly?

Source: paper p11, p12 (refs 30, 66)

10 · The boldest tool

Gene editing gave a hint, not proof.

Contested

Edited channel catfish had an FCR of 1.18 against 1.57 for controls. The gap was not statistically significant, because each group had two replicates.

In red sea bream, knocking out one gene gave about 9% better feed efficiency at an early stage.

What stands between a lab fish and a farm?

Source: paper p13 (refs 72, 73, 74)

11 · What it changes for you

Ask where the fish was tested.

A fish that wins in one tank can lose on another farm.

The review says to test efficiency on the diet and in the conditions the farm will really use.

How strong is the proof behind each tool?

Source: paper p9, p14, p15. Questions: vizmaya reading.

On what feed?
In what water?
How many fish were measured?

12 · The limits

Not every tool has the same proof behind it.

Many generations of breeding give the firmest results. Big gene hunts and DNA prediction are solid, but depend on how they are tested.

Most single-gene links, RNA signals, gut microbes and epigenetic marks show association, not cause. Gene editing is earliest of all.

So what slows everything down?

Source: vizmaya reading of paper p1, p9 to p13

13 · The limits

The slow part is counting what each fish eats.

To breed for feed use, you need each fish's own intake. Cameras, X-ray machines and single cages can do it, but slowly and at high cost.

VideoTilapia in groups, watched on camera.p3, ref 13
X-raySalmon smolts, feed tracked on X-ray images.p3, ref 14
Single cageTurbot, one fish per cage.p3, ref 12

Fewer measured fish means weaker studies and smaller training groups.

And what do the rules say about edited fish?

Source: paper p3, p14 (refs 89, 90)

14 · The limits

Gene-edited fish are treated differently by country.

Some places have cleared certain edited fish. Others still treat many edits like GMOs.

Who did the work, and what did we check?

Source: paper p13, p14 (refs 83, 84)

JapanGenome-edited fish have been commercialised.
ArgentinaNon-GMO decisions for genome-edited fish.
BrazilNon-GMO decisions for genome-edited fish.
EUStricter. Many edits fall under GMO rules.
USOversight depends on the change and its risks.
ChinaInside its farm-biotech framework. Pathways still developing.

Sources and method

Credit, and what we did not check.

vizmaya desk · 6 November 2026

Review by Usman Abdullahi, Ibrahim Kandeh, Jianlin Li, Yuanfeng Xu, Yongkai Tang and Yongfeng Zhao. Biology 2026, 15(20), 1849. Open access, CC BY.

Every number here comes from that review and was matched to its page. We did not open the 109 studies it cites. Their reading list is longer than ours.

Caveats on the slides: shrimp 0.743 and turbot 0.43 carry weaknesses, and the catfish result was not significant.

Sources

Photos, Wikimedia Commons