Most automated transfers are dense and regular: fill a whole plate, stamp one plate onto another, dilute every well the same way. Cherry-picking is the opposite. You want a handful of specific wells, scattered across one or several source plates, gathered into a consolidated destination, and which wells you want is decided by data that did not exist until this run. It is sparse, it is irregular, and it changes completely from one run to the next.
That variability is the whole character of the task, and it is also where the discipline lives. The routing changes every run. The way each individual transfer is performed should not. Hold those two apart and cherry-picking is manageable. Blur them and you get a process that is different every time in ways you cannot account for.
What cherry-picking actually is
Cherry-picking, or hit-picking, is selecting particular wells out of a larger set and moving just those into a new, denser layout, driven by a list rather than by geometry. The list comes from somewhere real. In a screen you are pulling the hits, the compounds that scored, out of thousands that did not. In a QC workflow you are pulling the failures, the samples that need to be redone. In consolidation you are gathering scattered samples off many partly used plates into one full one. The common thread is that the pattern of wells is arbitrary and specified per run, not baked into the method.
What drives the whole operation is the pick list, and its shape is simple: for each transfer, a source, a destination, and a volume. Source plate and well, destination plate and well, and how much to move. Everything the run does is an execution of that list, row by row.
Random access, one row at a time
A sparse, arbitrary list of transfers wants an instrument that can reach any single well independently, so cherry-picking is naturally a job for single channels or for a head with independent channel control. That random access is what lets you hit A1 on one source, then H12 on another, then something in the middle of a third, without the fixed geometry that an eight-channel or full-plate transfer imposes. You are trading the throughput of moving many wells at once for the freedom to move exactly the wells you want.
Because each row of the list is its own little transfer, the temptation is to think of them as unrelated events. They are not. They are the same operation repeated with different coordinates, and the physical performance of that operation, the aspirate, the dispense, the speeds and depths, is the part that should be identical across every pick. That constancy is what makes a hundred different transfers comparable to one another.
Tips and carryover
The stakes in cherry-picking are usually the samples themselves, which are often precious and often the very ones you flagged as important, so contamination between picks is the thing you most want to avoid. Tip strategy is the lever, and it is a real trade.
- Fresh tip per pick: a new tip for every transfer eliminates carryover completely and is the default for anything where cross-contamination is unacceptable, at the cost of tips and of the time spent changing them.
- Washing and reusing: washing a tip between picks saves consumables and time, but it only makes sense when you have proven the wash actually clears your liquid, and it is a poor bet for sticky, high-value, or trace samples.
When the samples are irreplaceable, a fresh tip per pick is rarely the wrong choice. Carryover in a cherry-pick does not just add noise; it can move a real hit into a well that should have been empty, or taint a redo with the very sample it was meant to replace.
Sources drain, decks fill up
Cherry-picking pulls repeatedly from source wells, and wells go down as you draw from them. If several picks come from the same well, or if a source is already low, the liquid level drops during the run and a fixed aspiration depth that started just below the surface ends up either sipping air or plunging to the bottom. Following the level, lowering the tip as the well drains, keeps the aspiration honest. It is easy to forget precisely because a full-plate transfer never has to think about it.
Then there is the physical reality of many source plates. Consolidation in particular may want to reach across more plates than the deck holds at once, which pulls in plate handling, sensible ordering of the run, and the plain limit of how many positions you have. A little planning pays off here.
- Order picks to reduce travel: grouping the list by source plate and by region cuts head movement and wall-clock time without changing a single volume.
- Order picks to reduce tip changes: where washing is allowed, clustering picks of the same or compatible samples lets you get more out of a tip before it has to change.
- Respect the deck footprint: if the sources outnumber the positions, the run has to be staged, and the staging has to be planned rather than discovered mid-run.
Identity is the deliverable
The point of a cherry-pick is not just that the right volumes moved. It is knowing, afterward and with certainty, which source landed in which destination. A consolidated plate of hits is worthless if you cannot say that destination well B4 came from source plate 3, well F11. Preserving that mapping, and carrying the sample identity through the transfer, is the actual product of the run. Everything else is plumbing in service of it.
Which is why the most damaging failure in cherry-picking is not a bad volume but a mismapped list: a destination that is off by one row, a source plate numbered wrong, a list that drifted out of sync with the plates physically on the deck. A mismap does not look like an error. It looks like a perfectly successful run of the wrong thing, and it can invalidate every conclusion drawn downstream. Verify the list against the real plates before you start, not after.
In cherry-picking the routing is different every run and the liquid class never is. The list decides where the sample goes; the class decides that it arrives intact and traceable.