How USDA Actually Measures a Nutrient — Inside the Lab Methods

USDA doesn't publish a nutrient value and leave the method to your imagination. This dataset documents 275 distinct analytical lab methods, linked to specific nutrients 1,005 times across 221 of its 474 tracked nutrients — from AOAC combustion assays for protein to ICP-MS for trace minerals. Here's what that actually looks like, nutrient by nutrient.

How many of this site's nutrients actually have a documented lab method?

This database tracks 474 distinct nutrients. Querying the project's own lab_method, lab_method_code and lab_method_nutrient tables turns up 275 uniquely named analytical methods, broken into 292 individual method-code records — USDA's internal shorthand, like SPGP_S for a gravimetric specific-gravity method or ICP_S for inductively coupled plasma — linked to nutrients through 1,005 separate method-to-nutrient records. Only 221 of the 474 nutrients tracked here, a little under half, have at least one named method attached. The rest are measured indirectly: calculated from other nutrients, summed from components, or carried forward without their own fresh lab chain of custody in this particular download. That split isn't a flaw in the data — it's exactly the distinction this site's own provenance panel is built to surface. On a Foundation Food page, the "How this was measured" panel lists the recognized analytical technique for every lab-analyzed nutrient, and says nothing where a value was calculated instead.

It's also worth noting the 292 method-code count is slightly higher than the 275 method count: USDA occasionally assigns more than one internal code to the same named method, typically to track a different sample-preparation variant under an otherwise identical analytical description. The method record stays the single source of truth for what technique was actually used; the code is closer to a lab's own bookkeeping shorthand for how a sample reached that method.

Do individual amino acids get their own documented methods too?

Yes, and they're some of the most heavily re-documented nutrients in the dataset. Alanine alone carries seven separate method records: an acid-hydrolysis HPLC technique cited by DOI, a second HPLC method referencing a 1988 Sawar et al. paper, two different acid-hydrolysis/performic-acid variants of AOAC 982.30, a microbiological assay (AOAC 960.47), an HPLC method from a 1986 USDA guidebook, and an ion-chromatography technique logged as "Beckman_AA." Arginine shows a near-identical spread. That density exists because amino acid analysis has gone through several generations of chemistry — microbiological assays were standard decades before ion-exchange and HPLC methods became routine — and this dataset keeps each cited method on record rather than quietly collapsing them into one undocumented "amino acid analysis" label.

What does a specific method actually look like — protein as an example?

Take protein. This dataset lists six separate method records for it, and every one uses the same underlying technique — combustion — while citing a different validated protocol: AOAC 968.06 + 992.15, AOAC 992.15 (39.1.16), AACC 46-30, and three more variants. That's not six competing ways to measure protein; it's six documented instances of labs citing the specific published AOAC (AOAC INTERNATIONAL, the standards body) or AACC method a given sample actually followed, rather than a vague "lab tested" label. On Broccoli, raw, for instance, this site's own food-nutrient table shows protein's derivation as "Calculated" rather than "Analytical" — meaning that particular value was derived from other measured components, not assayed directly, even though protein does have documented lab methods elsewhere in the dataset.

How does USDA test something trickier, like a trace mineral or a B-vitamin?

Selenium is a good contrast case: seven method records, spanning hydride-generation AAS, ICP, the newer ICP-MS protocol AOAC 2011.19, and even an isotope-dilution GC-MS technique cited by DOI. Selenium, Se on Broccoli's own page is marked "Analytical" — an actual lab-measured 1.6 micrograms per 100g, not a calculated estimate. Thiamin shows six methods split between fluorometric assays (AOAC 942.23) and HPLC runs, including one logged simply as "In-house HPLC thiamin." Vitamin C, total ascorbic acid has four: two microfluorometric methods under AOAC 967.22, a titrimetric method (AOAC 967.21), and an HPLC method cited by DOI. Each technique exists because different compounds need different chemistry — you can't run the same assay for a water-soluble vitamin that you'd run for a mineral.

What does "SR Legacy" mean when it shows up in a nutrient's Source column?

Ingredient pages built from USDA's Survey/FNDDS data carry a "Source" column next to every nutrient value, and it sometimes reads "SR Legacy" rather than "Foundation." On Tomatoes, grape, raw, most values — protein, energy, potassium, vitamin C — are sourced from "Foundation," meaning they're pulled straight from the newer Foundation Foods program. But the sugars figure on that same ingredient is sourced from "SR Legacy code 11529," meaning that specific number predates Foundation Foods and was carried over, unchanged, from the older Standard Reference database. According to USDA FoodData Central's own data type documentation, SR Legacy "was the primary food composition database in the United States for decades" and received its "final release April 2018" — it is no longer updated, so any value still sourced to it is, by definition, at least several years old even on a page published this year.

What do the "Analytical," "Calculated" and "Summed" derivation labels mean?

Every nutrient value on a Foundation Food page carries a derivation label, and this project's food_nutrient_derivation table resolves 79 specific codes down into 10 broader source categories — things like "Analytical or derived from analytical," "Calculated or imputed," and "Value manufacturer based label claim for added nutrients." A real example makes the distinction concrete: on the fortified milk record Milk, reduced fat, fluid, 2% milkfat, with added vitamin A and vitamin D, the Vitamin D3 value (1.1 micrograms) is marked "Analytical" — independently lab-measured — while the Vitamin A, RAE value on the same food is marked "Calculated," and the combined Vitamin D (D2+D3) figure is marked "Summed." Three different derivation labels, three different levels of lab involvement, on the same product.

Why do some nutrients have six documented methods and others have none?

Multiple method records usually mean a nutrient has been re-analyzed across different labs, decades, or revisions of a reference protocol — not that USDA can't settle on one technique. Protein's six combustion variants, logged above, are a clean example: AOAC updates and revises its numbered methods over time, and this dataset preserves which exact version a given measurement cited. By contrast, the 253 nutrients in this dataset with zero rows in lab_method_nutrient were never given that documentation in this particular download — some are simple sums of other nutrients (like total Vitamin D from D2 and D3), and some are tracked nutrients for which this dataset's import never carried method-level detail forward at all.

What should I actually do with this information?

When a Foundation Food's provenance panel lists a named technique for a nutrient, that's a direct signal the value was independently verified in a lab, not calculated or estimated. When it isn't listed, check the "Derivation" column in the nutrient table itself — "Analytical" still means lab-measured even without a named method attached to this particular build; "Calculated," "Summed," or "Imputed" mean the number was derived from other data instead. According to USDA's own Foundation Foods documentation, the program is built specifically to preserve "individual sample data points" and their acquisition metadata rather than publish a single composited number, which is the same transparency goal this site's provenance panel and lab-method tables are translating into plain language.

None of this is a reason to distrust a calculated value — plenty of nutrient figures, like total carbohydrate by difference, are supposed to be calculated rather than directly assayed. The practical payoff is being able to tell the two apart on sight: a named method or an "Analytical" derivation means a lab instrument produced that number for this specific food; anything else means it was built from other numbers already on the page. Next time a food's nutrient table shows a figure you want to lean on heavily — for a recipe calculation, a comparison, or just general curiosity — that single distinction is the fastest way to judge how directly it was measured. The site's methodology page covers this same provenance chain in more depth, beyond what fits on any single food's panel.