Five cross-cutting findings

In one paragraph. Five questions cut across every category: whether artificial intelligence changes the picture, whether blending proteins really raises their quality, which specific organisations in the region a Vietnamese actor could work with, what halal certification would require of an exporter, and what Vietnam's own traditional fermentation organisms are worth. None of them reorders the ranking. Two of them, the regional contract manufacturer and the halal requirement, change what a first plan should contain.

13.1 Does artificial intelligence change any of this?#

This was the last untested of the four claims in chapter 14. The finding is that the acceleration is real, narrow, and strongest in exactly the places where it is least discussed. [general]

• Strain and pathway engineering for precision fermentation. The clearest measured gains: fewer variants screened, faster design cycles, better performance from the best candidates, with laboratory validation showing large activity gains and parallel campaigns completed in about a week. This is a real, validated acceleration.
• Culture medium and growth factor optimisation. The best quantitative claim in the evidence set is 40 to 80 percent lower media cost using Bayesian optimisation or active learning, in proof-of-concept studies. Given that medium is the dominant running cost in cell culture, and growth factors are 90 to 95 percent of it, this is the highest-leverage application found.
• Protein structure prediction and design from scratch. Accuracy gains and fewer candidates to test are documented, with strong experimental validation of designed proteins, but public evidence with end-to-end before-and-after numbers for food proteins is thin.
• Scaffold design and bioprocess optimisation. Acceleration described in words only. No before-and-after numbers.
• Predicting what a protein will do in a food from its sequence or composition. Promising, evidence thin.

Honest summary. Where artificial intelligence has been measured in this field it saves experiments rather than years, and the single best-evidenced result is a reduction in media cost rather than a discovery breakthrough. Claims of end-to-end acceleration in alternative protein research rest on benchmarks rather than on production outcomes. This does not change the category ranking. It is an argument about how a small Vietnamese programme could get more out of a limited research budget, not a reason to prefer one protein category over another.

13.2 Protein complementation, which closes open question NP-G3#

Blending a cereal protein short of lysine with a legume protein short of sulfur amino acids does raise protein quality, and this is supported across reviews and animal feeding studies through improved amino acid balance, PDCAAS, digestibility, protein retention and growth. Wheat and bean, maize with cowpea or soy, rice and legume, and wheat and mung bean examples are all described. The gain is not automatic and depends on the ratio, the processing, and whether the blend is still short of one amino acid.

But the number Vietnam would need does not exist. The improvements were measured in various ways: rat feeding studies, laboratory digestibility, human or pig digestibility used to calculate DIAAS, and purely calculated blend scores. For the cereal and legume categories that matter here, the sources do not give a directly measured rice, wheat, maize or mung bean blend DIAAS of 100 or above. Blends reaching 100 were shown by calculation or simulation, not by feeding tests. [general] See contradiction CS-C2.

What that means. A Vietnamese rice-plus-legume product can be argued to have good protein quality on a calculated basis, which is what most of the industry does. It cannot yet be claimed on measured evidence. If the project wants that claim, someone has to measure it, and that is a defined, fundable piece of work.

13.3 Regional partners a Vietnamese actor could actually use#

Named, specific and checkable, which is what the brief asked for rather than country profiles. [VN-adjacent]

• ScaleUp Bio, Singapore. A precision fermentation contract development and manufacturing facility in Tuas, 2,300 square metres, with fermentation capacity up to 10,000 litres and thermal processing to 100 litres. It is a joint venture of Nurasa, which is Temasek's food innovation platform, and ADM. It received a food manufacturing licence from the Singapore Food Agency in May 2024, and has named customers from Australia (Nourish Ingredients), the United States (C16 Biosciences), Malaysia (Ultimeat) and Singapore (Allium Bio). It offers process optimisation, quality control and Asia market entry support alongside manufacturing. This is the single most actionable regional finding in the study: a licensed, food-grade, third-party fermentation facility of meaningful scale, in the same time zone, already working with Southeast Asian clients.
• Esco Aster, Singapore. Holds a Singapore food processing licence to manufacture cell-based cultivated meat and operates as a manufacturing platform for partner companies, working towards ISO 22000 and FSSC 22000 certification. Capacity figures are not published. Relevant to Vietnam only if the posture on cultivated meat changes.
• Singapore as the regional regulatory gateway. Singapore remains the only ASEAN jurisdiction with granted novel food approvals, with a microbial protein on sale since 2022 and a precision-fermented dairy protein since 2023, assessed in roughly 9 to 12 months once a complete dossier is submitted, according to an advocacy source rather than the regulator.

The strategic read. A Vietnamese venture does not need Vietnamese fermentation capacity in order to have a product. It needs a Vietnamese feedstock advantage, a strain, and a contract. That reorders what the first two years of any plan should contain. The enquiry to make is specific: minimum working volume, downstream recovery fees, and the tariff treatment of shipping a formulated ingredient back into ASEAN markets.

13.4 Lines that added constraints without changing the ranking#

Halal certification of fermentation-derived ingredients [general]#

Relevant only to the export framing, and there it is material.

• Halal review covers the whole system, not just the purified molecule: input sources, process controls, contamination prevention, traceability and audit documentation.
• Culture media are the main exposure. Animal-derived peptones, serum, growth factors and processing enzymes are high-risk inputs unless the source is clearly halal, traceable and often separately certified.
• Genetic modification is not itself disqualifying. The questions are whether the host and the inserted genes come from permitted sources, whether the process is safe, and whether forbidden inputs are used upstream or remain as residues.
• Alcohol is a control point even when none is added. Certifiers look at ethanol formed during fermentation. Indonesia is the most explicit, with published summaries pointing to a 0.5 percent ethanol threshold tied to MUI rulings.
• Segregation. A dedicated halal production line is the simplest route; shared lines can be certified with validated cleaning and documented segregation. This bears directly on using a third-party contract manufacturer, and should be raised with ScaleUp Bio in the same conversation as price.
• Precedent is thin. Accepted fermented precedents exist for vinegar and yeast in Indonesia, but recombinant-protein fermentation and cultivated systems lack settled, broadly accepted rulings.
• Malaysia is strong on plant inspection and halal assurance management; Indonesia is more explicit on traceable material status and alcohol; Gulf references are general and standards-based.

Implication. A Vietnamese fermentation product intended for Muslim-majority export markets should specify animal-free media from the first process design rather than retrofitting it, and should treat the halal status of a precision-fermented protein as an open question to resolve with a certifier early.

Indigenous strains from traditional ferments [VN-adjacent]#

Interesting, with one concrete lead, but not decision-changing.

• Tetragenococcus halophilus, a salt-tolerant lactic acid bacterium, recurs across fish sauce, soy sauce and salted seafood and anchors the isolate literature.
• The clearest documented industrial lead is a Vietnamese fish sauce Bacillus isolate producing a salt-stable protease.
• Thai thua nao soybean ferments have yielded Bacillus species with documented uses in protein hydrolysate production, enzymes, antimicrobials, clot-dissolving compounds, exopolysaccharide and poly-gamma-glutamic acid.
• Vietnamese men rượu and bánh men starters are mixed communities of moulds, yeasts and bacteria, including Amylomyces and Rhizopus-type moulds, Saccharomyces and lactic acid bacteria, which could be developed into defined starter cultures.

Why it does not change the ranking. These are enzyme and hydrolysate leads, not hosts for bulk protein production. They matter as a national biodiversity asset and as a story, and they connect to the OPENPichia finding in the sense that a strain nobody else controls is valuable. Before any of it is promoted, someone needs to check what Vietnam's biodiversity law and the Nagoya Protocol require of anyone commercialising Vietnamese strains. Open question CS-G1.

Straw mushroom and solid-state mycelium: a closed line#

• Vietnam produces 250,000 to 270,000 tonnes of mushrooms a year, ranks 11th globally in mushroom exports, and is simultaneously the third largest importer of processed mushrooms and second largest importer of mushroom powder. Straw and oyster mushrooms dominate, and the sector's stated constraints are poor spawn quality, limited research and development, fragmented manual production, and a shelf life of 5 to 10 days even refrigerated. The market was estimated at USD 362 million in 2025.
• Conversion is poor. A documented cycle at Tan Binh Cooperative in Thanh Binh district, Dong Thap: about 10 tonnes of rice straw in 500 bales produced 700 kilograms of mushrooms over a 45-day cycle, generating 7.4 million VND from mushrooms plus 4 million VND from selling the spent substrate as organic fertiliser. Adoption at that site was 5 to 7 households.
• Solid-state mycelium is not a food protein ingredient. The established food examples of mycelium grown by solid-state fermentation use edible substrates such as soy, cereals or food side streams. For straw and similar woody residues, published work treats the fermented material as feed, as a mycelium-plus-substrate composite, or as mushroom substrate, not as a harvested human food protein. No like-for-like comparison of protein yield per tonne against submerged fermentation exists in the sources.

Verdict. Straw mushroom is a valuable smallholder livelihood and a genuine circular-economy story, and it demonstrates that Vietnamese rice straw can be collected and biologically converted. As an industrial protein route it is far inferior to submerged fermentation on conversion, cycle time and labour. The project should cite it as context and not pursue it. The one live question it leaves is why a country producing 250,000 tonnes of mushrooms imports processed mushroom and mushroom powder at that rank, which is a processing and shelf-life problem rather than a protein problem. Open question CS-G2.

Purple non-sulfur bacteria: attractive story, weak numbers [general], with a [VN-direct] hook#

• Biomass on real agro-industrial and food-processing wastewater ranges from below 0.5 to about 4 to 5 grams per litre. Protein is around 40 to 55 percent on some food-processing wastewaters, and higher for selected Rhodopseudomonas cultures, but high protein content does not track high productivity.
• The biomass carries carotenoids and vitamins, though species-level carotenoid values and vitamin B12 content are not reported in the evidence set.
• Selection for these organisms is strongest under light and airless conditions. Closed photobioreactors achieve this but are far too expensive for a low-value protein, so the field is moving towards open raceway designs with controlled timing of oxygen, organic carbon and darkness.
• Pilot scale only. No large commercial deployment, no species-specific aquaculture feeding trial outcomes, and no cost per kilogram for any process.
• The Vietnamese hook: the very first literature run in this project found 28 photosynthetic bacterial isolates from coastal Huế, four of them at 36.25 to 50.75 percent protein, published in the Hue University Journal of Science in 2025. Vietnam has indigenous strains already characterised for exactly this.

Verdict. Biomass concentrations of 0.5 to 5 grams per litre compare with 8 to 118 grams per litre for the fungal systems in this study. At the extremes that is two orders of magnitude, and between the upper bounds it is less than a factor of two, so the honest statement is that the productivity range sits well below the fungal range without being uniformly so. Either way, treat it as a wastewater treatment technology that yields a saleable by-product, which is a legitimate and locally useful thing, rather than as a protein production route. The Huế isolates are worth citing as evidence of domestic capability, not as a business case.

Methanotrophic protein on biogas: unresolved, with a hypothesis worth testing#

This line did not complete. No organisms, biomass concentrations, costs, scales, gas transfer data or explosion-risk findings were retrieved, and nothing from this line is used anywhere in this report.

An external reviewer argues the line should be abandoned rather than re-run, on two grounds: aerobic methane fermentation requires handling methane and oxygen mixtures within narrow flammability limits, which mandates expensive explosion-rated infrastructure, and Vietnam's distributed agricultural biogas digesters are small, remote and loaded with corrosive hydrogen sulfide. Both points are plausible and neither is sourced. They are recorded here as the leading hypothesis for why this route is hard in Vietnam, and as exactly what a re-run should test.

The line remains worth running, because it failed as a tool error rather than returning a negative result, because the Vietnamese logic is good on its face, and because one literature search costs nothing. Vietnam has a large distributed base of pig-farm biogas plants and offshore gas, and methanotroph protein is one of the few routes with genuine commercial precedent. Open question F-G10.