Selected yeasts

Where it is practiced

The practice comes from elsewhere. Oenology and brewing have been inoculating for over a century. Coffee has only done so since the 2010s.

The commercial benchmark exists, but it is weak. Lallemand, a Canadian company from oenology and brewing, sells a range of yeasts for coffee under the name Lalcafé, including the strains Oro and Cima. The date of 2014 for the launch of Oro comes from the manufacturer's communication. I could not cross-check it: the Lallemand website refuses automated consultation. I therefore do not say that Oro was the first yeast sold for coffee, being unable to verify it.

The published trials that I was able to read in full all come from Brazil. The best-documented lot was fermented at the Chuá farm, near Patos de Minas, in the Cerrado Mineiro. The farm is at 1,050 m altitude, at 18° 35′ 15′′ south latitude and 46° 25′ 35′′ west longitude. Two varieties were followed there, Arara and Paraíso. The genetic study of the tank covered only Arara. The work on isolating and selecting strains comes mainly from Lavras and Uberlândia.

A detail that matters for what follows: the strain used in this trial, RX60, is a Laffort product, not a Lallemand product. The commercial benchmark of the sheet and the best-documented trial do not point to the same supplier.

In Colombia, Cenicafé describes the addition of microorganisms in August 2023, in its Avance Técnico 554. The center lists it among the modified fermentation methods, but did not include it in its own trial. The six treatments it compared in Cesar were:

• fermentation in the fruit before pulping;
• fermentation on blends of pulped coffee;
• fermentation with added water;
• fermentation with added coffee pulp;
• fermentation without oxygen;
• the control, run with the Fermaestro.

None was an inoculation treatment.

Elsewhere, the practice exists among microlot producers on every continent, since yeasts have been sold everywhere since the mid-2010s. No census says how much coffee is actually inoculated.

The principle

A spontaneous fermentation starts with what is there. The flora arrives by several doors:

• the skin of the fruit;
• the soil and the process water;
• the machines;
• the hands.

It changes from one farm to another, and from one day to another on the same farm. Inoculating means refusing this lottery. A chosen strain is introduced, in large numbers, from hour zero.

The bet is a population bet. A yeast that arrives alone must first multiply before it weighs in. A yeast that arrives by the billions works at once, and deprives the others of sugar and space.

What it does to coffee comes down to two chemistries.

The first is enzymatic. The yeast secretes pectinases, enzymes that cut the pectin chains of the mucilage, the layer of sugary flesh stuck to the parchment. Cenicafé names four: polygalacturonase, pectinesterase, pectin lyase and pectate lyase. They release galacturonic acid, the acidic sugar that forms the backbone of pectin, and the gel comes loose. A reservation is needed here. The fruit has its own pectinases, and the bacteria of the tank produce them too. The degradation of the mucilage is therefore never the sole work of the added yeast.

The second is alcoholic fermentation proper. The yeast consumes the glucose and fructose of the mucilage through glycolysis, then converts pyruvate into ethanol and carbon dioxide. One molecule of glucose gives two molecules of ethanol and two of carbon dioxide. In parallel, it makes esters, including ethyl acetate. It also makes higher alcohols from amino acids, by the Ehrlich pathway: the amino acid is transaminated, decarboxylated, then reduced to alcohol. Leucine thus gives isoamyl alcohol.

These metabolites do not stay in the tank. They diffuse through the parchment into the bean, and some survive roasting. This is how fermentation reaches the cup.

The bet, for its part, is not always won. Molecular monitoring shows that the introduced strain does not necessarily dominate.

The sequence

1. Choose the strain. The three names that come up most in the literature I have read are Saccharomyces cerevisiae, Candida parapsilosis and Torulaspora delbrueckii. A team from Lavras published four selection criteria:

• grow between pH 2.0 and pH 8.0;
• tolerate osmotic pressure;
• tolerate heat;
• tolerate the accumulation of its own metabolites.

These four criteria tell what a coffee tank is. The medium is acidic and sugary, it warms up, and it gradually becomes toxic to its occupant. Saccharomyces cerevisiae tolerates a pH of 2.3 to 8.6, which explains its place.

2. Wake the yeast and build up the inoculum. Published protocol: 3 g of freeze-dried yeast in 100 mL of sterile medium, in a 250 mL flask. The medium contains, per liter:

• 6 g of yeast extract;
• 10 g of sucrose;
• 45 g of KH2PO4;
• 1 g of MgSO4 7H2O;
• 2 g of (NH4)2SO4;
• plus 30 g/hL of Superstart, a yeast activator sold by Laffort.

The whole is adjusted to pH 5.5. Incubation at 30 °C under shaking at 100 rpm up to 10^9 cells/mL, counted in a Neubauer cell. This rich medium and this heat serve only one purpose: to raise the population before the encounter with the coffee.

3. Harvest and prepare. Ripe cherries, mechanically washed just after harvest, with artesian well water in the Brazilian trial. Depending on the route, the fruit is pulped or kept whole.

4. Load the tank. In the Brazilian trial: 200 L high-density polyethylene bioreactor, about 124 kg of cherries, lid fitted with a bubbler valve. The valve lets carbon dioxide out and prevents air from coming in. Nobody drives out the air present at the start: the microorganisms consume it, and the medium becomes anoxic by exhaustion, not by purge. This is what is called SIAF, self-induced anaerobic fermentation.

5. Inoculate. 10% of the volume as inoculum. Submerged variant: 60 L of water added, that is, 30% of the total volume. Cenicafé recommends the same ratio for its submerged fermentations, 30 L of clean water per 100 kg of slimy coffee, because a liquid substrate is more homogeneous than a solid substrate.

6. Hold the temperature. The Brazilian trial did not settle on a single setpoint. Arara was run in a closed chamber at 27 °C, Paraíso at 18 °C, each time against tanks left at ambient temperature in the shade. Cenicafé for its part compared two constant setpoints, 15 and 20 °C, with a night-day cycle of 18 to 26 °C, or 23 °C on average. It was at 15 °C that the frequency of specialty and premium quality cups went up. Temperature governs speed, and speed decides what remains to be consumed. Refrigerated between 4 and 8 °C, a mucilage had degraded only 20% of its reducing sugars at 31 hours. At 23 °C on average, it had degraded 30% in the same time.

7. Monitor. In the Brazilian trial, 150 g samples are taken in triplicate at hour zero, then every 24 hours. Cenicafé recommends measuring the pH and degrees Brix of the mucilage at the start and at the end. The falling Brix and the falling pH say the population is working.

8. Stop. The durations tested in Brazil are 24, 48 and 72 hours. Cenicafé gives tighter benchmarks, indexed to temperature. In an open solid tank, 14 to 18 hours between 20 and 23 °C, and 14 to 24 hours between 13 and 17 °C. In a submerged tank, 18 to 30 hours between 20 and 23 °C, and up to 42 hours between 13 and 17 °C. The center notes that defects become more frequent beyond 42 hours. Beyond that, the risk is no longer insufficiency but excess.

9. Wash, dry, rest. In the Brazilian trial, the coffee is washed with the same well water as before fermentation. It then dries in the sun on asphalt terraces, to 11 to 12% moisture. It is stored at room temperature and hulled after 30 days. This rest serves to even out the moisture in the lot before hulling. Cenicafé for its part recommends a rest of the dry parchment coffee of at least fifteen days, between 18 and 20 °C.

Seeding with something other than a pure strain: the case of the SCOBY

This sheet describes inoculation as the addition of an identified strain, bought in dry or liquid form and rehydrated. There is a practice that departs from it, documented on a named project, and it deserves its place because it sheds light on the border between inoculating and subculturing.

Kornman reports that the Catracha project, in Honduras, used a kombucha SCOBY to seed its fermentations. The acronym stands for symbiotic culture of bacteria and yeast, that is, the gelatinous disc that forms on the surface of a fermented sweet tea and that houses yeasts and acetic bacteria side by side in a cellulose matrix they make themselves. It is therefore not a strain, it is a community, and a community whose exact composition varies from one disc to another and from one house to another.

The interest of this case is that it occupies an intermediate position between the two practices this sheet sets against each other. On one side spontaneous fermentation, where the population forms by chance from whatever arrives with the cherries. On the other the inoculation of an identified strain, which imposes a known population. The SCOBY imposes a population, so it is not spontaneous, but this population is not characterized, so it does not offer the traceability that justifies inoculation. It brings dominance without bringing knowledge.

A remark on mechanism is needed, and it is not trivial for coffee: a kombucha SCOBY contains acetic bacteria in abundance, since it is to them that a kombucha owes its characteristic acidity. Seeding a tank with such a disc is therefore to deliberately introduce the organisms that most directed processes seek to eliminate. This makes sense only in an aerobic or partly aerated context, and it assumes the producer is looking precisely for the acetic note of ripe fruit rather than fearing it. No measurement, no control and no protocol are given by the source, which is a booklet without bibliography: the mention counts as a testimony of a practice, not as a result.

The same page also notes that suppliers are actively developing coffee-specific strains, which is consistent with the genera used listed elsewhere, Pichia, Saccharomyces and Wickerhamomyces. The direction of the trade is thus the reverse of the SCOBY: from the community to the strain, and from the strain to the strain selected on the targeted substrate. The Catracha case is interesting as an avowed counterexample, not as a trend.

Sources for this passage: Kornman, p. 18. Genera used as cultures for coffee: Hackler and Särchen, chapter 4, p. 49.

What it gives in the cup

What is measured.

The Chuá farm trial had its tanks scored by five certified Q Graders. The protocol is that of the Specialty Coffee Association: 8.25 g of coffee for 150 mL of water, medium roast on a single profile. The control did not ferment at all: the washed cherries went straight to drying.

A point of design must be read before the scores. On Arara, twenty-four tanks were set up and only twelve received the yeast Saccharomyces cerevisiae RX60. The other twelve fermented spontaneously. The design crosses three other factors: solid or submerged tank, held or ambient temperature, and 24, 48 or 72 hours. A difference in the cup therefore never relates to the yeast alone.

The scores published from this trial go in that direction. On Paraíso, the unfermented control scores 83.60 ± 0.89 and the best tank 86.05 ± 0.57, a gap of 2.45 points. But three tanks score below the control: 82.50, 82.80 and 83.10. On Arara, the control scores 82.65 ± 1.63 and only two tanks rise significantly, by 3.60 and 3.65 points. The others cannot be distinguished from the control. The 2025 article that publishes the sensory part of this trial gives the same orders of magnitude. It announces about 2.5 points of gain at best under the 18 °C setpoint, and about 3.5 points under the 27 °C setpoint.

The lesson of this table is not that inoculation improves. It is that management decides. And a methodological reservation must be added: the control was not fermented at all, so the gap measures fermentation and inoculation together, never the yeast alone.

The molecular monitoring of the same lots, published in 2026, is harder still. The RX60 strain did not establish itself. It was detected only at low abundance, and the genus Saccharomyces appears only in the inoculated tanks. This proves in passing that it does not belong to the natural flora of these cherries. The indigenous yeast Kazachstania, for its part, occupied 83.6% of the fungal community in the inoculated tanks, against 85.4% in the spontaneous tanks. The authors speak of a low competitiveness of the exogenous strain against the local flora. They put forward two probable causes, without demonstrating them: the progressive acidification of the medium, and the rapid multiplication of better-adapted indigenous yeasts. The inoculated communities differed from the control, but coexisted with the native flora without suppressing it.

One gain, only one, is measured without ambiguity: the inoculated replicates varied less among themselves. Regularity is therefore demonstrated, but on microbial composition, not on the cup score.

The 2025 trial on four varieties compares two strains in the wet process, 48 hours between 20 and 24 °C, in triplicate. The two strains are not of the same species, and the sheet must say so: CA11 is a Saccharomyces cerevisiae, B10 is a Kluyveromyces lactis. The second had never been used as a starter culture on coffee before this trial. All the scores exceed 83, and the combination Paraíso 2 with CA11 exceeds 85. But the total scores do not differ significantly between treatments, at the threshold p > 0.05. The score difference is therefore not a result of this trial.

What is a result is the chemistry. One hundred volatile compounds were counted in thirteen classes, and two of them appear only in the inoculated lots: 5-methyl-2-furfuryl alcohol and n-hexadecanoic acid. This is not an intensification of what existed, it is the appearance of something absent from the controls.

The descriptors call for the same caution. The floral descriptor appeared only in the seeded lots, with one strain or the other. The fruity notes, for their part, were present everywhere, including in spontaneous fermentation: they therefore distinguish nothing. With B10, two varieties out of four received a descriptor set of their own:

• citrus;
• lemongrass;
• honeyed lemon;
• yellow fruit.

Three results are reported by others, without my having read the original text. The figures are those that the 2026 article attributes to them:

• Jimenez et al., Food Bioscience, 2023: Candida parapsilosis in the dry process, 2.15 points more than a fermentation with indigenous flora alone;
• Janne Carvalho Ferreira et al., LWT, 2024: co-inoculation of Saccharomyces cerevisiae and Bacillus amyloliquefaciens in a submerged tank, 86.75 points;
• Mota et al.: Saccharomyces cerevisiae, up to 5 points of gain in the dry process and in pulped natural.

What is claimed.

Sellers promise predictability, faster kinetics and reproducible profiles. The predictability part is supported by the variability measurement cited above. The descriptor lists attached to each strain, on the other hand, are commercial discourse. I found no published and readable blind comparison that establishes them strain by strain.

The scientific literature nevertheless attributes characters to the species. These lists come from literature reviews that I could not trace back to the original trial:

• Saccharomyces cerevisiae: citrus, caramelized sugar, banana, cashew;
• Torulaspora delbrueckii: milky and cocoa notes;
• Meyerozyma caribbica: caramelized sugar;
• Candida parapsilosis: caramelized sugar, citric acid, apple, cherry, winey character.

What goes wrong

The first risk is specific to the method, and it is measured: the inoculum does not take. The introduced strain remains a minority against the indigenous yeasts. The producer has paid for the yeast, believes he is running a directed fermentation, and is in fact running a largely spontaneous one. Nothing in the tank tells him so.

The second risk is that of any long fermentation: excess. An insufficient or excessive fermentation produces abnormal amounts of acetic acid, butanoic acid or propionic acid, which give an undesirable taste and smell. The defect has a name in the Colombian trade, fermented (fermento, in Spanish), and its neighbor is called vinegary (vinagre). The monthly surveys of Almacafé on coffee in warehouses give the order of magnitude. In September 2013, 5.8% of cups were defective, of which 34% fermented and 23% reposo (in Spanish, coffee left moist too long before drying). In July 2014, of 3.5% of defects, fermented weighed 37%. In December 2014, of 2.3% of defects, fermented and vinegary weighed 41.3%. Fermented is therefore the first or second defect of the country depending on the month.

A broader survey confirms the ranking. Of 580 samples of parchment coffee taken from 162 farms in seven departments, harvests 2005 to 2007, 36.8% of cups had a defect. Fermented affected 11.9% of samples and accounted for 32.3% of defects.

The third risk is reposado, coffee left moist too long before drying, and the fourth mohoso (moldy, in Spanish). Cenicafé groups these four defects together in its prevention guide: fermento, reposado, fenólico and mohoso.

The enterobacteria that proliferate in a badly run tank produce named compounds: 3-isopropyl-2-methoxy-5-methylpyrazine, 2,3-butanediol and butyric acid. These are off-flavors, false tastes, not aromas.

A measured side effect deserves to be noted without being over-interpreted. In the 2026 trial, the inoculated tanks favored two filamentous fungi, Cladosporium and Fusarium, less abundant in the spontaneous tanks. The authors do not link this finding to a cup defect. I record it as a measurement, not as proof of harm.

Finally, enriching the substrate with sugar or pulp to boost the microbial load aggravates everything above. Cenicafé writes that this route demands rigorous procedures and complementary infrastructure, and that particular care must be taken with the concentrations used.

There remains the regulatory risk, and I must state it in the conditional. Three claims circulate. The Cup of Excellence would forbid any foreign addition while allowing yeasts and mosto. The Specialty Coffee Association would allow any addition prior to the green coffee stage. The Best of Panama would have disqualified four infused coffees. I could verify none of the three in a published rulebook. What remains certain is that the line between seeding and infusion moves from one competition to another and from one year to another. A producer aiming at a competition must read the year's rules, not a sheet.

Benchmarks

• Lalcafé range from Lallemandlaunch of Oro announced in 2014 by the manufacturer, not cross-checked
• Strain of the best-documented trialSaccharomyces cerevisiae RX60, Laffort
• Dose published in this trial10% of the volume, inoculum built up to 10^9 cells/mL
• Tank200 L high-density polyethylene bioreactor, about 124 kg of cherries
• Durations tested in Brazil24, 48 and 72 hours
• Run temperatures of the Brazilian trial27 °C on Arara, 18 °C on Paraíso
• Setpoints compared by Cenicafé15 °C, 20 °C, and a night-day cycle of 18 to 26 °C
• Share of the indigenous yeast Kazachstania after inoculation83.6%, against 85.4% in spontaneous fermentation
• pH tolerated by Saccharomyces cerevisiaefrom 2.3 to 8.6