NRAS Q61H: a pre-registered negative, and the test that would settle it

NRAS is a molecular switch that cells use to tell themselves to grow. It turns itself off by cutting the last piece off its own fuel, and a single water molecule is the blade that does the cutting. A mutation called Q61H, found in several cancers including colorectal, jams that off-switch, so the growth signal stays on.

The short version

We wanted to test a specific idea: that the mutation does not change the chemistry of the cut, but how the water is aimed at the target. We ran eight careful computer simulations to look for it.

They came back uninformative. Our main, pre-registered measurement — how often a properly lined-up, ready-to-cut water sat in position — was essentially zero in every run, in both the healthy and the mutant protein. That is a limitation of the simulation method, not a discovery about the mutant. One side-observation looked interesting but fell apart the moment we pressed on it.

We are publishing the negative anyway, for three reasons: so nobody else burns the same compute, to be honest about what our pre-registration actually found, and to point at the one cheap, tractable wet-lab experiment that nobody has ever run and that would actually settle this.

Where we stand, in plain tiers: the disease is certain, the mechanism is likely, the fix is a guess.

The idea we tested

NRAS carries a fuel molecule (GTP). To switch off, it snips the last phosphate group off that fuel, with help from a partner protein. The chemical scissors is a water molecule that attacks the phosphate. For that attack to work, the water has to be held in a precise orientation.

This points to a cleaner hypothesis. Isotope-effect work on Ras GTP hydrolysis — Du & Sprang (2009, Biochemistry 48:4538, PMID 19610677), using oxygen-18 kinetic isotope effects — characterized the transition state of the cut as “loose,” and did not pin the chemistry of the cut on the residue-61 side chain. So the hypothesis worth testing is that Q61H changes how the water is positioned and aimed, not the chemistry itself. That is what we set out to test.

What we actually did

Eight GPU molecular-dynamics simulations, 100 nanoseconds each: four of the healthy protein and four of the Q61H mutant, each with the partner protein present. Total compute cost: $52.35.

We pre-registered our primary measurement before looking at any results: the fraction of simulation time that a competent water — one lined up in true attack geometry — sat in the right place.

What came back

The primary result was uninformative. The competent-water fraction was essentially zero across all eight runs, healthy and mutant alike. The honest reading is that today’s classical simulations and water models cannot hold that reactive geometry at all, so the measurement cannot tell the two proteins apart. This is a tool failure, not a disproof of the hypothesis and not a finding about the mutant.

The one lead — and why we don’t trust it

One side-observation looked promising: the healthy protein seemed to keep a water in the catalytic seat far more often than the mutant did. We treat this as exploratory and post-hoc, not a claim, and here is why it does not hold up:

• Two of the four mutant “empty seat” results happened before the real simulation even started — the water had already drifted away during setup — and both came from the same starting structure.
• If we counted any of three known crystal waters as filling the seat, the difference vanished entirely.
• Each simulation froze into a single shape and barely changed over its whole run, so four runs are not four independent measurements.

The statistic behind that lead was a four-versus-four test giving p = 0.014 — and that number is simply the smallest possible value at that sample size, not strong evidence of anything.

The follow-up checks

Everything else we looked at was negative. The position of residue 61 did not drive water alignment. Crowding in the pocket split evenly. Twenty separate measurements at the leaving-group end showed nothing.

A real-world hint that a base can help

We are not reasoning in a vacuum. A 2026 study in Nature Chemical Biology (RCSB PDB entry 12XE; Wang, Zhang et al., “Brønsted-basic small molecules activate GTP hydrolysis in KRAS-Q61 mutants”) solved a crystal structure of KRAS carrying a Q61 mutation, bound to a non-hydrolysable GTP analogue together with a drug-like molecule that carries a Brønsted-basic methyl-imidazole group. The authors report that such molecules can switch GTP cutting back on in KRAS Q61 mutants — direct experimental support for the idea that supplying a base near the water can help restore the cut.

We state the limits plainly: that structure is the Q61K mutant, not Q61H; it is KRAS, not NRAS (the catalytic machinery is identical across the RAS family, but it is a different protein); and it is a structure-plus-activity result, not a full mechanism. It strengthens our case. It does not prove it.

The experiment nobody has run

Here is the genuinely open, publishable gap. To our knowledge, no one has ever measured a pH-rate profile on any RAS Q61 mutant — the one measurement that would show directly whether an introduced histidine is acting as an acid, a base, or neither.

The context makes it worth doing. The only pH-rate data we are aware of on the healthy protein is flat: Mistou, Cool & Parmeggiani (1992, Eur. J. Biochem. 204:179, PMID 1740128) found that the GTP-cutting activity of H-ras is unaffected by pH across the range 6 to 9.2. Against that flat baseline, any pH dependence introduced by the Q61H histidine would stand out cleanly.

The precedent for the readout comes from a related enzyme. On EF-Tu, Maracci et al. (2014, PNAS 111:14418, PMID 25246550) combined pH-rate profiles, solvent isotope effects and mutagenesis and concluded that a histidine (His84) helps position the attacking water. Q61H introduces a histidine into the RAS pocket. That is a direct, testable parallel.

The experiment: measure the pH-rate profile of GTP hydrolysis for Q61H. If the cutting rate rises as pH goes up, the histidine is acting as a base; if it falls, it is acting in its charged form; if it is flat, the histidine is not gating the chemistry at all. Every outcome is informative, and nobody has the data.

The best part: Q61H protein is available off the shelf (Reaction Biology, about $630). This is a cheap, fast measurement, not a multi-year program.

Why we are posting a negative

Because pre-registration only means something if you report what it found. Our simulation could not settle the aiming hypothesis, and saying so openly saves anyone else the same wasted compute. More usefully, it sharpens the one measurement that would move this forward — and we would rather a wet lab run it than have it sit undone.

Honest odds: a publishable mechanism paper, if the experiment is run, is maybe a 35–40% shot. An actual drug reaching a patient from this line of work is low single digits. The disease is certain, the mechanism is likely, the fix is a guess — and we will keep saying exactly that.

If you run rapid-kinetics or pH-rate work on small GTPases and want to collaborate on this measurement, get in touch.