Your 260/280 is measuring your buffer

I am a spectroscopist, not a molecular biologist. My own peer-reviewed work is in microwave spectroscopy, which means I have spent a large share of my career on a narrow question: what is actually responsible for the number an absorbance measurement gives you. That question travels. A ratio of two absorbances taken at two wavelengths behaves the same way whether the sample is a gas-phase rotational transition or a tube of RNA, and the failure modes are the same failure modes.

So this is not a report about whether the 260/280 ratio is useful. It is useful. It is a report about what it responds to, which is a longer list than the convention implies, and about the fact that the list is documented in plain language by the companies that make the instruments. Nobody is hiding this. It just does not survive the trip into lab protocols, where a threshold gets written down and the caveats do not.

The convention

The rule everyone learns is short. From Thermo Fisher's NanoDrop technical bulletin:

A ratio of ~1.8 is generally accepted as "pure" for DNA; a ratio of ~2.0 is generally accepted as "pure" for RNA.

The same bulletin adds that if the ratio is appreciably lower, "it may indicate the presence of protein, phenol or other contaminants that absorb strongly at or near 280 nm."

Hold on to two things there. The reference values are 0.2 apart. And the diagnosis attached to a low ratio is already a list of three different things, not one.

Artifact one: the solution you measured in

In 1997, William Wilfinger, Karol Mackey and Piotr Chomczynski published a study in BioTechniques on exactly this. Their abstract states the finding without hedging:

Adjusting the pH of water used for spectrophotometric analysis from approximately 5.4 to a slightly alkaline pH of 7.5-8.5 significantly increased RNA A260/280 ratios from approximately 1.5 to 2.0.

Read that against the convention. An RNA prep that reads 1.5 looks bad enough to re-extract. The same RNA reads 2.0, textbook pure, if you dilute it in slightly alkaline water instead of slightly acidic water. The sample did not change. The solvent did.

The authors also report that the effect is not pH alone: "changes in both the pH and ionic strength of the spectrophotometric solution influenced the A260/280 ratios." And they observed the underlying spectral behavior directly, noting that the 260 nm absorbance maximum seen in water "was shifted by 2 nm to a lower wavelength when determinations were carried out in Na2HPO4 buffer at a pH of 8.5."

Thermo Fisher cites this paper in its own documentation and states the magnitude as a rule of thumb: "Acidic solutions will under-represent the 260/280 ratio by 0.2-0.3, while a basic solution will over-represent the ratio by 0.2-0.3."

That single artifact is larger than the entire gap between the DNA reference value and the RNA reference value.

Artifact two: which instrument you used

This one is pure spectroscopy and it is my favorite, because it comes from the shape of the curve rather than from anything chemical.

Thermo's bulletin explains it in two sentences: "Although the absorbance of a nucleic acid at 260 nm is generally on a plateau, the absorbance curve at 280 nm is quite steeply sloped. A slight shift in wavelength accuracy will have a large effect on 260/280 ratios."

That asymmetry is the whole problem. Your numerator sits on a flat part of the spectrum, so a small wavelength error barely touches it. Your denominator sits on a slope, so the same small error moves it a lot. You are not dividing two equally-behaved quantities; you are dividing a stable one by a twitchy one, and every wavelength error lands entirely on the twitchy one.

The documented magnitude:

For example, a +/- 1 nm shift in wavelength accuracy will result in a +/- 0.2 change in the 260/280 ratio. Since many spectrophotometers claim a 1 nm accuracy specification, it is possible to see as much as a 0.4 difference in the 260/280 ratio when measuring the same nucleic acid sample on two spectrophotometers that are both within wavelength accuracy specification.

Two instruments. Both in spec. Both correctly calibrated. Same tube. Answers 0.4 apart, which is twice the distance between "pure DNA" and "pure RNA."

This is the practical reason that comparing a ratio measured today against one a colleague measured on a different instrument two years ago is not a comparison. It is also the reason a lab that switches platforms will see its historical purity numbers shift, and will usually blame the samples.

Artifact three: what you extracted

The third one gets left out of protocols almost entirely, and it undercuts the idea of a universal threshold more thoroughly than the other two.

The five bases do not share a 260/280 ratio. Thermo's bulletin lists the values estimated for each nucleotide measured independently: guanine 1.15, adenine 4.50, cytosine 1.51, uracil 4.00, thymine 1.47. The document then says what follows from that:

The resultant 260:280 ratio for the nucleic acid being studied will be approximately equal to the weighted average of the 260/280 ratios for the four nucleotides present. It is important to note that the generally accepted ratios of 1.8 and 2.0 for DNA and RNA respectively, are "rules of thumb". The actual ratio will depend on the composition of the nucleic acid.

A GC-rich template and an AT-rich template have genuinely different correct answers, because guanine at 1.15 and adenine at 4.50 are not interchangeable inputs to an average. The bulletin even explains the headline convention with this: RNA reads higher than DNA in large part because uracil, at 4.00, sits where thymine, at 1.47, sits in DNA.

Which reframes the whole convention. 1.8 and 2.0 are not thresholds of purity. They are typical weighted averages for typical base compositions, measured under typical conditions, and the document that gives them says so in the same breath.

Adding it up

The specification is 0.2 wide. Solution pH is worth 0.2 to 0.3. Instrument wavelength accuracy is worth up to 0.4 between two compliant machines. Base composition sets a different baseline per template. Any one of those is enough to move a sample across the line, and in a real lab they are all acting at once, in unknown directions.

This does not make the number worthless. It makes it a screening tool with a wide tolerance rather than a purity certificate with a sharp edge. A ratio of 1.2 is telling you something real. A ratio of 1.75 versus 1.85 is telling you nothing you can act on.

What the ratio actually detects

Here is the other half of the honesty, and again it is in the manufacturer's own text: "Abnormal 260/280 ratios usually indicate that the sample is either contaminated by protein or a reagent such as phenol or that there was an issue with the measurement."

Three possibilities from one number, and they call for three different responses. The ratio is a nonspecific alarm. It says something absorbs near 280 nm, or something went wrong. It does not say which, and it cannot.

The bulletin is blunter still about the converse case, which is the one that costs people experiments:

It is important to note that there are occasions when the purity ratios are within expected limits, yet there is a problem with the sample.

And it names what the number is standing in for: "the best indicator of DNA or RNA quality is functionality in the downstream application of interest."

That is the sentence I would put on the wall. The purity ratio is a cheap proxy for a question you actually care about, which is whether the sample will work in the assay. Proxies are fine. Proxies mistaken for the thing itself are how a bad sample gets a passing grade and a fine sample gets thrown away, which is the same substitution I have written about here with detection limits and with error bars.

The 260/230 companion, and one trap in it

The second ratio most people record is 260/230, expected in the range 2.0 to 2.2. A low value points at contaminants absorbing near 230 nm, and the documented list is specific: carbohydrate carryover, which is a common problem with plant material; residual phenol; residual guanidine from column-based kits; and glycogen used for precipitation.

A high 260/230 is more often a measurement problem than a sample problem, and the two named causes are a dirty pedestal and an inappropriate blank. The bulletin's rule for the blank is worth repeating exactly, because it is the same lesson as artifact one: "The blank solution should be the same pH and of a similar ionic strength as the sample solution." Blanking in water and measuring in TE is a documented way to produce low 260/230 values on clean samples.

The detail I would flag hardest is about extraction chemistry. Guanidine hydrochloride, used in DNA isolations, absorbs at around 230 nm, where the 260/230 ratio will catch it. Guanidine isothiocyanate, used in RNA isolations, absorbs at around 260 nm. Carryover of the RNA-kit reagent lands on the same wavelength you are using to quantify RNA. It inflates your apparent concentration and it does not announce itself in the ratio designed to catch reagent carryover. If you have ever loaded what you thought was plenty of RNA and got a weak result, that is a mechanism worth knowing about, and it matters before library prep, not after.

What to do instead

None of this requires new equipment. It requires four habits.

Standardize the solution and say what it was. Wilfinger and colleagues end their abstract with their own recommendation: they "found RNA A260/280 ratios to be more reliable and reproducible when these spectrophotometric measurements were performed at pH 8.0-8.5 in 1-3 mM Na2HPO4 buffer." Note that this fixes two things at once. The same abstract reports that "the ability to detect protein contamination was significantly improved when RNA was spectrophotometrically analyzed in an alkaline solution." The condition that stabilizes the number also makes it more sensitive to the thing it is supposed to detect.

Blank in the same solution the sample is in, at the same pH and similar ionic strength. This is one sentence in a bulletin and it removes an entire class of phantom results.

Look at the spectrum, not only the ratio. The shape carries information the quotient throws away: the peak should be at 260 nm and the trough at about 230 nm, and a contaminant will move both. A ratio compresses a curve into one number, and the diagnostic content is in the curve.

Do not compare ratios across instruments, and record which instrument produced each one. Given artifact two, a purity value without an instrument attached is an incomplete record.

And treat the number as a gate, not a grade. Reject the clearly bad. Do not rank the ambiguous.

What I did not verify

I read the Wilfinger, Mackey and Chomczynski abstract in full through Europe PMC and every quotation from it above is verbatim from that abstract. I did not obtain the full 1997 article, which is not open access. Nothing here characterizes their methods or their data beyond what the abstract states.

Two of my three sources are manufacturer technical bulletins rather than peer-reviewed papers, and I want to be exact about what that supports. I am not citing them as independent scientific findings. I am citing them as documentation of what the instrument maker states about its own instruments and about this measurement, which is precisely the claim I am making: the caveats are published, by the vendor, in the documents that ship with the technique. Where the pH magnitude is concerned, the bulletins cite Wilfinger and colleagues as their source, and I have gone to that paper's abstract directly rather than relying on the restatement.

The per-nucleotide 260/280 values are attributed in bulletin T042 to Lehninger's Biochemistry, 2nd edition, 1975. I did not open that book, so those five values are reported here as the bulletin reports them.

One small oddity worth naming for anyone who goes to the source. Technical Note 52646 lists a cause of a low 260/280 ratio as "A very low concentration (>10 ng/µL) of nucleic acid." The surrounding text describes a very low concentration, so the inequality appears to be inverted in the published bulletin. I have not quoted that item as a load-bearing claim, and I mention it only so that nobody reproduces the typo believing it is a threshold.

Finally, the numbers here are documented magnitudes from specific documents, not universal constants. If a newer revision of either bulletin changes them, this report needs a dated update and will get one.