Turning heads, frames, shells and trim into a seasoning base: protease grades by activity, a second de-bittering enzyme, odour control and a bench-trial grid for fish and shrimp by-product.
Request pricingHeads, frames, shells and trim are worth more as a seasoning base than as meal, provided the base is not bitter and the room does not stink. Fish sauce is the oldest version of the step: fish and salt left to break down over time, with the breakdown itself as the slow part. Traditional makers add no enzyme because they do not have to: a whole small fish such as an anchovy or a sardine brings its own gut enzymes to the barrel. Gutted trim, frames and shells from a filleting or peeling line do not, which is why a by-product base needs a bought protease and why the choice of protease decides whether the result is savoury or bitter.
Sellers of blended flavour protease to oyster-sauce and fish-sauce makers describe one shared problem: bitter peptides lower palatability and long fermentation delays production. Their claimed result is bitter peptides largely degraded, umami clearly up and the production cycle cut by about 50 per cent. Treat the cycle figure as a claim to test on your own trim, not as a plan. The same pair of failures, low extraction and bitterness that has to be controlled during hydrolysis, is what the bovine-bone collagen peptide blends are built around, with an alkaline protease as the base of the blend. A single protease cut deep enough to lift yield releases the bitter fragments; the second enzyme that trims those fragments is what the blend adds.
A process comparison for collagen, soy and keratin peptides puts enzymatic hydrolysis at 92 per cent or better amino-acid retention against 75 per cent for acid or alkali routes, a directional enzyme blend at a 98 per cent or better molecular-weight pass rate, and optimised pretreatment at a 20 per cent cut in production cycle. Hydrolysis time depends on the substrate: 4-6 hours for collagen, 3-5 hours for soy and 8-12 hours for keratin, with a target molecular weight below 1000 Da. The by-product hydrolysis blend sold for poultry, bone by-products and fish, shrimp and clam is pitched on exactly that point: meat flavours, bone broth and seafood extracts without the harmful by-products of acid or alkali hydrolysis.
The animal-protein hydrolysis blend, made up of a protease, an exopeptidase and a flavour enzyme, is specified for a working temperature of 30-60 C and a 12 month shelf life. Its suggested dose is 0.1-0.3 per cent of substrate, the same band the tenderising papain for fish and seafood carries, in both cases with the note that the real figure depends on production conditions. The sea cucumber peptide hydrolase, also a blend with an exopeptidase and a flavour enzyme, is tighter: working pH 7.0-8.0, working temperature 54-56 C, suggested dose 0.3-0.5 per cent, sold in a 1 kg bag. Its claimed results are a degree of hydrolysis up to eighty per cent or more, effective protein use above seventy-five per cent, a pale colour, a clear solution and no bitterness. The eighty is a best case, and a sea cucumber window is a starting point for shrimp heads, not a specification. None of these tables states salt tolerance, water-to-solids ratio or the stop step; the bench trial sets those.
Alkaline protease for seafood by-product comes in two grades of the same Bacillus licheniformis enzyme: a 200,000 U/g grade sold for chicken, duck, fish, beef, shrimp and crab hydrolysis, and a 100,000 U/g grade sold specifically for fish and shrimp off-cuts. An animal protease marketed for liver, pork, beef, fish, shrimp meal and blood meal hydrolysis is also listed at 100,000 U/g. For the same enzyme load, expect the 200,000 U/g grade to need about half the weight of the 100,000 U/g grade.
The product sold for fish-protein flavour boosting and peptide de-bittering is not more alkaline protease; it is an aminopeptidase at 5,000 U/g. Aminopeptidase is an exopeptidase: it clips amino acids off the ends of the peptides the first protease leaves behind, which is where the bitterness sits. The blends above build that step in, which is why they list an exopeptidase and a flavour enzyme beside the protease. Chasing bitterness with a bigger dose of the first protease is the usual mistake: it cuts deeper, makes more short bitter peptides and costs more enzyme.
Trim and offal carry off-notes that survive salt, and a dedicated deodorising protease grade is now sold for offal and by-product streams with gamey or urine-like odours. Protease grades are also sold by sector: an alkaline-stable grade for detergents, a tenderiser and soy-sauce grade for food plants, and an enteric-coated grade for digestive aids, so the first question to a supplier is which sector grade you are being offered. Skin and scales are a separate stream from flesh trim: a collagen-peptide enzyme is sold specifically for cod skin and tilapia scales, and that stream deserves its own bench trial rather than a share of the head kettle.
A base that comes out bitter still has a buyer. Aquafeed mills short of fish meal use enzyme-hydrolysed protein at 0.1 per cent soy hydrolysate for shrimp and 2-4 per cent peptide blend for carp, and report growth up 5-15 per cent and feed digestibility up 20-30 per cent. Small-scale producers turn fish processing waste into a semi-moist feed and a liquid fertiliser, and the questions under that training film come from small farmers asking for affordable feed, with one asking how to stabilise fish hydrolysate with phosphoric acid.
| Route | Amino-acid retention | Bitterness | Cycle | Where it fails |
|---|---|---|---|---|
| Acid or alkali hydrolysis | 75 per cent | high | long | loses about a quarter of the amino acids before the kettle |
| Single alkaline protease, 100,000 or 200,000 U/g | 92 per cent or better | high when cut deep for yield | 4-6 hours on collagen-type substrate | bitter base when pushed for extraction |
| Protease with exopeptidase and flavour enzyme, 0.1-0.3 per cent at 30-60 C | 92 per cent or better | claimed largely degraded | claimed about 50 per cent shorter | unverified until trialled on your trim |
| Sea cucumber peptide hydrolase blend, 0.3-0.5 per cent, pH 7.0-8.0, 54-56 C | not stated | claimed none | not stated | eighty per cent degree of hydrolysis is a best case, not a day-one result |
| Second-step aminopeptidase, 5,000 U/g | unchanged | de-bittering is its job | adds a stage | wasted if the first cut was wrong |
| Cell | Enzyme | Dose, per cent of substrate | Temperature | What to log |
|---|---|---|---|---|
| A1-A3 | alkaline protease 100,000 U/g | bottom, middle and top of the 0.1-0.3 band | one setting inside 30-60 C, held by probe | free amino nitrogen, bitterness score, odour score, hours to target |
| B1-B3 | alkaline protease 200,000 U/g | the same three points, so the enzyme load is double row A at each dose | same setting as row A | the same four, plus salt level in the kettle |
| C1-C3 | protease with exopeptidase and flavour enzyme blend | the same three points | same setting | the same four, plus cycle time against the about 50 per cent claim |
| D | aminopeptidase 5,000 U/g run as a second step on the best A and B cells | set in trial | same setting | bitterness score before and after |
Running trim from a filleting line like a traditional fish-sauce barrel and waiting for enzymes that are not there, because the guts went out with the offal. Going acid or alkali to save enzyme cost and losing a quarter of the amino acids. Raising the first protease dose to fix bitterness instead of adding the 5,000 U/g aminopeptidase step. Reading the eighty per cent degree of hydrolysis as a specification rather than a best case. Buying the cheaper 100,000 U/g grade and dosing it like the 200,000 U/g grade. Putting cod skin and scales into the head kettle instead of their own collagen trial.
Email us for a bench-trial plan and samples of two enzyme grades: send the trim type, salt level, current cycle time and the complaint (bitter, smell or slow), and we reply with the grid above filled in for your stream.
Tell us your application and volume — we reply with pricing and lead time.