Gluten Quality: Several Measurements, and Why Protein Content Is Not One of Them

Flour bags advertise protein content because it is cheap to measure and easy to print. It is also, taken alone, a poor predictor of how a dough will behave. The properties bakers care about, extensibility, elasticity, tolerance to long fermentation, come from which proteins are present and how they are assembled, and there is a standard toolkit for measuring that.

Two protein groups, two jobs

Gluten is not one substance. The functional division is clean and worth committing to memory: gliadins contribute extensibility and viscosity, glutenins contribute elasticity and strength.

Gliadins account for 40 to 50 per cent of total flour protein and come in classes distinguished by molecular weight: alpha at 25 to 35 kilodaltons, beta at 30 to 35, gamma at 35 to 40, and omega at 55 to 75. They are monomeric, which is why they lubricate rather than reinforce.

Glutenins are polymeric, and within them the high molecular weight subunits carry disproportionate influence. They make up only 1 to 1.7 per cent of flour dry weight and 10 to 12 per cent of total grain protein. Yet a 2024 review reports that they account for 45 to 70 per cent of the variation in baking performance among European wheats.

That ratio is the central fact about gluten quality. A little over one per cent of the flour explains most of the difference between a strong flour and a weak one.

Each cultivar expresses only 3 to 5 of these subunits, and one gene, Glu-A1y, is silent in all cultivated tetraploid and hexaploid wheats. So the genetic space is small, discrete, and identifiable, which is what makes allele-based quality prediction possible at all.

What the standard tests measure

A 2026 study of eighteen hard red spring wheat samples gives a useful picture of the range within a single high-quality market class.

Wet gluten ran from 31.12 to 36.80 per cent. Gluten index, which measures what fraction of the wet gluten resists being forced through a sieve and therefore indexes strength rather than quantity, ran from 81.31 to 96.32. Farinograph water absorption ran from 61.35 to 65.15 per cent, consistent with hard red spring wheats generally exceeding 60.

The protein composition measures in the same study show what underlies those numbers. Unextractable polymeric protein, expressed as a percentage, ran 56.34 to 60.94. The glutenin to gliadin ratio ran 0.65 to 0.86. High molecular weight glutenin as a share of total glutenin ran 40.88 to 44.61 per cent, and the ratio of high to low molecular weight subunits ran 0.69 to 0.81.

Unextractable polymeric protein deserves a note, because it is the most mechanistically informative of these. It reflects how much of the glutenin has polymerised into aggregates too large to extract, which is close to a direct measurement of the network the dough will form.

How well composition predicts performance

The same study tested predictive models, and the results are a caution against reductionism.

A model using protein quantity together with starch and fibre achieved an R-squared of 0.862. A model using protein composition measured by reverse-phase HPLC achieved 0.842. A model using size-exclusion HPLC achieved 0.733.

The bulk composition model, in other words, performed marginally better than the detailed protein composition model, and both beat the polymer size distribution model. Detailed protein chemistry buys real explanatory power, but it does not straightforwardly beat simpler measurements within a single well-defined wheat class.

Where allele scoring breaks down

The Glu-1 quality score assigns numerical values to high molecular weight glutenin alleles and sums them, on the reasoning that better alleles should give stronger dough. A 2024 study tested that against two measurements and found the relationship real but not orderly.

Zeleny sedimentation rose with score in the expected direction, from 40.8 millilitres at a score of 5 to 49.2 at a score of 10. Alveograph deformation energy, W, ranged from 206.6 to 250.6 in units of ten to the minus four joules across scores, and the study describes this relationship as non-monotonic. Higher score did not reliably mean higher W.

The allele-level data show why. Sorting Glu-B1 allelic pairs by measured W gives 251.6 for 17 plus 18, 236.5 for 7 plus 9, 230.8 for 7 plus 8, 184.8 for 6 plus 8 and 166.7 for 7 alone. Across the whole set, W ran from a minimum of 135.8 to a maximum of 251.6. Specific allele combinations matter more than their summed scores, which is what one would expect if the subunits interact rather than contribute independently.

The sedimentation tests are not equivalent

Sedimentation tests are cheap, fast, and widely used for screening, and there are two common variants. A 2022 study compared them across the same material.

SDS sedimentation values ranged from 15.0 to 70.0 millilitres, averaging 38.6, with a coefficient of variation of 45.7. Acetic acid sedimentation ranged from 18.5 to 53.5, averaging 32.5, with a coefficient of variation of 23.6.

The SDS test spreads samples out roughly twice as much, which for a screening tool is exactly what you want. Its correlations with functional traits were moderate and in sensible directions: 0.69 with bread porosity, 0.64 with dough development time, 0.62 with overall evaluation, minus 0.71 with a dough deformation index, and minus 0.45 with dough softening.

Correlations of that size are useful for ranking many samples and inadequate for predicting one. A sedimentation number tells a breeder which lines to keep. It does not tell a baker how a particular sack will handle.

Reading a specification sheet

If you can get a full specification from a mill, the informative fields are these.

Gluten index, not just wet gluten, because the pair separates quantity from strength. Two flours at 33 per cent wet gluten with indices of 82 and 96 are different flours.

Farinograph water absorption, because it sets your starting hydration and because it is measured rather than guessed.

Alveograph W where available, remembering the range above: figures in the 130s indicate weak dough and figures around 250 indicate strong, and the difference is nearly twofold.

What will not help is protein content alone. It appears in the best-performing model above only in combination with other terms, and everything distinctive about gluten quality lives in the roughly one per cent of the flour that no bag has ever reported.