Troubleshooting an assay that worked last month

Last month the standard curve was linear and the blank was quiet. This month the same written method sits below the lowest standard, or the duplicates have opened into a fan. Nothing in that sentence tells you the cause. It tells you that at least one input moved. This selection guide is for choosing the first variable worth testing, not for declaring a winner from a hunch.

The inputs that most often hide inside "the same method" are the water, the pH at the temperature of use, the salt hydrate, the electrode slope, growth in a stored carboy, carbon dioxide, detergent left on glass, and the enzyme lot. Which contaminant a water meter can and cannot see is laid out in laboratory water types and where they fail. How a slope and a temperature belong in a buffer record is in preparing a buffer and checking pH. Replacement salts come from the reagents and chemicals catalogue, and a clean vessel from beakers and flasks. If the investigation ends in a repurchase, specify the hydrate on the quotation request.

Select what changed before you select a theory

Write a short inventory of the month. New enzyme lot, new bottle of a hydrate, new person at the pH meter, water system serviced, carboy left filled over a long weekend, dishwasher detergent changed, room hotter, meter rebooted after a power cut, buffer left uncapped. The first test belongs to an item on that list. A variable that did not change is a weak first test, however famous it is as a failure mode.

If you still have last month's buffer, water aliquot, or enzyme tube, the clean experiment is a single swap into an otherwise frozen recipe. One tube crosses old enzyme with new buffer. Another crosses new enzyme with old buffer. You are not running a screen of the whole laboratory. You are asking which swap restores the control. If nothing was retained, you cannot prove history. You can only remake the best-specified reagent and see whether today becomes consistent. Consistency is a weaker claim than "this was the cause", and the notebook should say so.

Do not start by changing eight things. A rescued assay after a heroic afternoon does not identify a cause, and next month you will not know which hero to repeat.

A worked month for a colour assay

Picture a plate assay that reads an enzyme rate. In September the slope of the standard was accepted. In October the rates are low, the blank is slightly higher, and the room has been hot. The inventory says: the enzyme lot is the same tube, the buffer was remade last week from a new bottle of the phosphate hydrate, the water came from a carboy filled two days ago, and the pH meter was last calibrated before a power cut. That inventory already ranks the work.

The electrode slope is the first measurement, because it is fast and because every pH claim depends on it. Calibrate with standards that bracket the buffer, as NIST's calibration notes remind a laboratory to do with real standards rather than with hope. If the slope is outside the laboratory window, stop. You do not yet know whether the buffer is wrong. Fix the meter, then read the buffer at the temperature the assay uses. A pass here does not prove the hydrate. It only returns pH to the list of things you are allowed to believe.

Next, read the new bottle. If September used an anhydrous phosphate and October used a hydrate, or the reverse, the molarity moved by the water of crystallisation even though the gram weight in the notebook was copied forward. Recalculate the mass from the formula weight printed on the bottle you actually opened. Remake once. If the control returns, you have a cause that fits the inventory. If it does not, the hydrate was a fair first chemical test and it is now retired.

The carboy is next if the blank is dirty or the buffer smells or looks hazy. Resistivity on a polisher upstream of that carboy does not audit two days of storage. Take water fresh at the point of use into a rinsed flask, remake a small buffer, and rerun the blank and one control. If the fresh water restores the blank and the carboy water does not, you have selected a storage problem. If both blanks stay high, look at the vessel.

Detergent is the vessel branch. Foam when you swirl, a sweet smell, or a flask that went through a washer used for general glassware is enough reason to rerun in glass or plastic dedicated to the assay, rinsed as the method says. Detergent can inhibit an enzyme and can also scatter light. Those are different readouts of the same residue. A rinse that still foams is not a rinse.

Carbon dioxide is the branch for an alkaline or weakly buffered reagent that sat open, and for ultrapure water stored in a half-empty bottle. The pH drifts, resistivity falls, and a blank that depends on either one moves. Recap, remake from a fresh dispense, and read pH without leaving the beaker open beside a busy bench.

The enzyme lot is last in this particular month because the tube did not change. It would have been first if a new lot had been opened on the day the assay fell and the old tube was still in the freezer. Run that old tube once before you email a complaint. If the old tube fails too, the lot is not the selected cause.

Match the symptom to a variable that can produce it

Public method collections such as Addgene and protocols.io show the same habit in different assays: a control that holds one reagent still while another changes. Use them as patterns for experimental logic. Do not copy their buffer into your assay because the logic was sound.

Branches that keep the selection honest

If the fresh-water buffer restores the control, retire the carboy contents and find out why the carboy sat. Do not pour the old water back "to finish the bottle". If the fresh-water buffer fails in the same way, water is no longer the first variable. Move to hydrate, pH temperature, detergent, then lot.

If two operators disagree on the pH of the same beaker, you have selected the meter, not the assay. One slope is lying. Agree on standards, storage solution, and the laboratory slope window before you remake ten litres.

If the assay is nucleic-acid work and the blank is a false product, water and the enzyme tube are both plausible, and detergent is less likely unless the extraction came through washed glass. Select by the control that is already in the method: no-template, no-enzyme, and a retained positive. The positive that fails while the no-template stays clean points at enzyme or at a missing cofactor, not at a dirty blank.

If everything passes in a small remake and fails only at full scale, select mixing, temperature gradients in a large bottle, and a precipitate that a small aliquot never grew. Large-scale failure is still one variable at a time. It is not a reason to change the method.

Heat, humidity, and an instrument that rebooted

A hot bench moves Tris pH by a few hundredths of a unit per degree. An assay set up in a warmer month is not the assay you validated in a cooler one unless the pH was set at the temperature of use. Humidity corrupts a slow weigh of a hygroscopic hydrate, so the "same mass" is fewer moles. A power cut clears some meter calibrations. After the power returns, the slope check moves to the front of the queue even if you were about to blame the enzyme. None of these is a claim about how often assays fail. They are reasons to pick a first test that the week's conditions made likely.

Safety

You are handling the same chemicals as the assay, sometimes in a more concentrated remake. Use the safety data sheet and the laboratory rules for acids, bases, and any biological material in the samples. A troubleshooting aliquot of an infectious sample is still an infectious sample. This page does not authorise a workaround, and it is not a diagnostic procedure for a clinical result. If the control fails, you withhold the result. You do not average it into a passing batch.

What the next purchase needs to say

When the selected cause is a reagent, buy the thing you actually tested. Name the salt form and hydrate, the enzyme unit definition if you are replacing an enzyme, the water grade if you are replacing water, and the storage temperature. Ask for the certificate line you will check on receipt. The quotation request should be specific enough that a different hydrate cannot arrive under the same nickname. On receipt, repeat the one control that failed, not the entire project, and file the result beside the lot identifier.