If your work touches a compound library, you work in DMSO whether you think about it or not. It is the near-universal solvent for small-molecule stocks, the fluid in your mother plates and your daughter plates, the medium every serial dilution for screening runs through. It earns that place because it dissolves almost anything and stays liquid over a useful range. And it repays you by being one of the hardest liquids on the deck to pipette accurately, for reasons that have nothing to do with each other and everything to do with why a compound stock quietly drifts off value over a long run.
What makes DMSO genuinely hard is that its problems compound. It is viscous, it is thirsty for water, and it freezes at a temperature you routinely cross. Each of those alone would justify a carefully tuned class. Together, in the small transfer volumes that screening demands, they make DMSO a liquid where a class that passed validation on Monday can be delivering the wrong volume by Friday without a single error message.
Viscosity: the familiar half of the problem
DMSO is noticeably more viscous than water, and that part behaves the way any viscous-liquid class does. Thick liquid flows into and out of a tip slowly, so a speed tuned for an aqueous buffer will under-aspirate because the liquid simply has not finished moving when the plunger stops. The fixes are the standard ones: slow the aspiration so the liquid keeps up, add a settling delay after aspiration so the meniscus and the film stabilize before the tip moves, and pre-wet the tip so the first transfer is not systematically drier than the rest. This is the tractable half of DMSO. If viscosity were the whole story, you would tune it once and move on.
The catch is that viscosity is not a fixed number. It depends strongly on temperature, and it depends on how much water the DMSO has taken up, which is exactly where the trouble starts.
Hygroscopy: the silent half
DMSO is strongly hygroscopic. Left open to the air, it pulls atmospheric moisture fast, and that water uptake changes everything you tuned your class against. As water accumulates, the viscosity drops, so a class calibrated for dry DMSO starts over-aspirating relative to its target. The density shifts, so if you validated gravimetrically your conversion from mass to volume is now slightly wrong. And most dangerous of all, added water can push dissolved compounds toward their solubility limit, so a stock that was clear can start to precipitate, taking your actual delivered dose down with it.
The reason this is the silent half is that it happens on the timescale of a run, not an instant. You open a source plate, the method runs for an hour or three, and over that window the DMSO in the exposed wells is slowly hydrating. Nothing alarms. The pipetting looks identical. But a class validated on fresh, dry DMSO is drifting the entire time, and the transfers late in the run are not the transfers you calibrated. This is why environmental control is not optional housekeeping for DMSO work. Low humidity in the room, sealed plates with the foil pulled only when needed, and a real awareness of the water content of your solvent are part of the method, not around it.
Freeze-thaw: the damage between runs
DMSO freezes at around eighteen degrees C, which sits right between your cold storage and your bench, so every stock you pull from the freezer or the fridge has to be thawed, and every stock you put back refreezes. Freeze-thaw is not free. Repeated cycles let the solution stratify, so concentration is no longer uniform top to bottom in a well, and they stress dissolved compounds in ways that degrade some of them over time. A library that has been cycled a dozen times is not the library the assay was designed against, and the pipetting is the least of it.
There is a handling consequence too. Because the freezing point is so close to room conditions, the viscosity of a stock coming off cold storage is markedly higher than the same stock after it has equilibrated to deck temperature. A class that aspirates correctly from a warmed plate will under-deliver from a cold one. Either let the plates reach a known temperature before pipetting, or accept that your class is only valid at the temperature you tuned it at, and hold that temperature.
Small volumes magnify all of it
Screening lives at small transfer volumes, often well below a microliter, and small volumes take every problem above and multiply it. The viscosity error is a larger fraction of a tiny dose. Evaporation, which DMSO does noticeably at low volume, removes a meaningful slice of a nanoliter-scale droplet. And precipitation from water uptake is far more likely to clog a fine tip when the passages are small. It is no accident that non-contact acoustic dispensing has become common for DMSO libraries, since firing droplets off the surface with no tip sidesteps the clogging and the contact carryover, though it trades those for its own dependence on a known, consistent solvent composition.
The one failure mode to fear
If you take one thing from all of this, make it the shape of how DMSO fails, because it is not the loud kind.
- Silent volume drift: as water accumulates in an open source plate over a run, the viscosity and density move and the delivered volume drifts away from the calibrated target, with no clog, no alarm, and no visible sign, so plates dispensed early and late in the same run are not truly the same.
That is the one to fear because it does not announce itself. A clog you can see, a precipitate you can spot, a cold plate you can warm. Hydration drift just happens, quietly, and it undermines the assumption a compound library rests on, which is that every well holds the concentration the plate map says it does. Guarding against it is mostly about controlling the environment and respecting the clock, and about knowing that a DMSO class carries an expiry the moment a plate is unsealed.
DMSO does not fail loudly. It hydrates, softens, and drifts while every dispense still looks perfect, which is exactly why a class tuned on dry solvent cannot be trusted to hold across an open run.