Scaling from pilot to screening

The pilot worked. Six wells, clean separation, a convincing figure, and now somebody wants it in 384-well format across a compound library by the end of the quarter.

The instinct is to multiply. Six wells became 384, so multiply the cells by sixty-four and place the order. That arithmetic is wrong in both directions at once, it over-buys cells because the wells are far smaller, and it under-buys almost everything else because scale introduces costs that did not exist at six wells.

This page is what actually changes. The short version: the cell requirement does not scale with well count, it scales with total growth area, and the growth area of a 384-well plate is disputed between two major manufacturers by 50 percent. Everything else that breaks is a consequence of the plate being smaller and the run being longer.

The arithmetic that people get wrong

Start with what a scale-up actually does to the cell order. The comparison below is a full plate in each format, at a single seeding density, using Corning’s published growth areas.

A full 384-well plate needs fewer cells than a full 6-well plate. Sixty-four times the wells at less than 40 percent of the cells. That is the whole reason screening formats exist, and it is the opposite of what the naive multiplication predicts.

The correct scaling rule: cells scale with total growth area, not with well count. Multiply your per-well area by your well count, multiply by density, and only then convert to vials.

The manufacturer disagreement that will change your order

Do not take the growth area from memory, and do not assume two manufacturers agree. On two formats they do not, and one of those two is the format you are scaling into.

A 50 percent disagreement on the format you are scaling into is not a rounding question. Use the published figure for the plate on your bench, from the manufacturer who made it. Corning’s own caveat is worth repeating: these areas are calculated from engineering drawings, do not account for moulding variation, and do not account for cells growing up the sides of the vessel.

The complete reference tables, including permeable supports and MEA formats, are in vials to wells.

What breaks: nine things, in order of how expensive they are

1. Volume tiering, which decides the budget before the vendor choice does

This is the largest single lever in a scale-up and it is available to you for free by asking.

Two lessons.

Cells tier hardest. A 43 to 46 percent fall in price per million between a 1M and a 5M vial, inside a single supplier, is larger than the spread between suppliers. Before you conclude a vendor is expensive, check you are comparing the same pack size. Size the whole campaign first, then price it.

Do not assume tiering always runs the same way. The last row is a real, live-verified counter-example: the larger AIM Biotech plate is more expensive per chip than the smaller one. Tiering is a commercial choice, not a law, and you have to check it per item.

2. The edge effect, which is a screening artefact and not a pilot one

At six wells you probably did not have an edge problem, because every well was an edge well and you were not comparing across positions. At 384 you are, and the outer ring evaporates faster than the interior over a multi-day run.

This is documented rather than folkloric: a published paper describes the edge effect as a global problem in 96-well plate culture. It produces a systematic position artefact that looks exactly like a weak compound effect, and it gets worse as well volume falls, which is precisely the direction you are scaling.

Three mitigations, in order of preference:

• Randomise treatment position across the plate, so the position effect becomes noise rather than bias.
• Leave the perimeter as buffer, filled with medium or PBS and excluded from analysis. This costs you 76 of 384 wells, which is 20 percent of the plate, a real cost, honestly stated.
• Use a humidified secondary enclosure where the platform supports one.

Whichever you choose, state it in the method. A screen with an unmitigated, unmentioned edge effect is a screen whose hit list contains a plate map.

3. Dead volume and dispensing overage, which grow with format

You cannot pipette the last of a suspension, and multichannel or automated dispensing across a full plate needs a reservoir that will never be fully recovered.

At six wells an under-estimate costs one well. At 384 it costs the plate, because a plate seeded in two halves from two different cell preparations is not one experiment.

4. Lot boundaries, which become unavoidable at scale

At pilot scale one vial covered the experiment. At screening scale the order may exceed what one lot can supply, and two lots of the same catalogue number can differ enough to break a comparison.

The published literature on this is direct: guidelines on addressing variability in iPSC-derived models exist because line-to-line and batch-to-batch variability is a reproducibility problem in this field, not a hypothetical one.

The rule: reserve the whole quantity from one lot and get it stated on the order confirmation. If a single lot genuinely cannot cover it, design the boundary in deliberately, split it across replicates rather than across compounds or conditions, and record which wells came from which lot. That is recoverable. Discovering the boundary afterwards is not.

For a campaign that will run across months rather than weeks, ask about a single-lot assay bank: material reserved and, in some cases, plated as assay-ready aliquots. No provider we surveyed publishes a price for that service. The only published empty-versus-assay-ready pair we could find is AIM Biotech’s, the idenTx 40 Plate empty at $444.00 per plate against AngioPrime, the same plate pre-loaded with hTERT-immortalised HUVECs, at $1,470.00, an uplift of $1,026.00 or 231 percent. That is one data point in one category and it is not a market rate, but it is the right order of magnitude for what pre-seeding costs. See /catalog/assay-ready-cell-plating.

5. Plate lead times, which are longer than cell lead times

At pilot scale you had the plates. At screening scale you need many of them, in a specific format, possibly with a specialist bottom or coating.

Buy the repeat plate with the first one. On any format with a multi-week lead, running out mid-campaign does not cost you a plate, it costs you the campaign’s schedule.

6. Instrument compatibility, which is binary

A plate that fits your imager may not fit your dispenser, and a plate format your platform lists may not be the format your specific instrument model reads. Bottom thickness, skirt height, well geometry and plate footprint all matter, and none of them are negotiable at the point of use.

Check the compatibility before ordering, not after. See plate format and instrument compatibility, and for electrophysiology specifically, the MEA platform compatibility matrix.

7. Assay quality metrics, which the pilot did not need

A six-well pilot is judged by eye. A screen is judged by a statistic, and the standard is Z-prime, a dimensionless measure of separation between positive and negative controls that accounts for both the signal window and the variability, defined by Zhang, Chung and Oldenburg in 1999.

Two consequences for the purchase order:

• Controls consume wells on every plate. Budget them into the well count from the start, not as an afterthought that displaces compounds.
• A positive control is not optional. Without a treatment known to move your endpoint in your hands, a null result is uninterpretable and the screen has no evidence it could have detected anything.

8. Compound and reagent consumption, which scales with wells, not area

The cells scale down with area. The compound, the medium changes, the assay reagent and the liquid-handling consumables scale with well count and run duration, which both go up.

This is the reversal that catches budgets. Cells become a smaller share of the total as you scale, and everything else becomes a larger one. A cost model built on the pilot’s cost structure will be wrong in its proportions even if the total happens to land close.

9. The failure rate, which is now attached to a much bigger object

At six wells a failure costs six wells. At 384 it costs a plate, the cells in it, the compound dispensed into it, the instrument time, and the schedule slot. The economics of a failed batch are set out in the cost of a failed batch; the point here is that the same failure probability applied to a larger unit is a larger number, and it belongs in the plan.

Practical consequence: buy for one more plate than you need. On a campaign with any repeat risk, the marginal cost of the spare plate’s cells at a volume tier is far less than the cost of re-ordering at small-pack pricing under time pressure.

MODELLED cost comparison across formats

This is a model. Every unit price is real, sourced and dated. Every quantity is our assumption. It is not a quotation.

The experiment: a cardiomyocyte viability screen at the FUJIFILM CDI published density of 1.56 × 10⁵ viable cells/cm², on Corning growth areas, one full plate in each format, cells purchased at BPS Bioscience’s published tiers.

Read the last two rows together. The same 384-well experiment costs $1,150 or $2,300 in cells depending on whose growth area figure you use. That is the 50 percent manufacturer disagreement propagating straight into a purchase order, and it is entirely avoidable by looking up the figure for the plate you bought.

Read the cost-per-well column downward. It falls by more than a hundredfold from 6-well to 384-well. That is the case for screening formats, and it is real. What it omits, deliberately, because we could not verify unit prices from named sources, is media, plates, coating, compound, assay reagent, liquid handling consumables, instrument time and labour, all of which scale with well count rather than with area. The cell line gets cheaper per well as you scale; the rest does not.

The staged scale-up that actually works

Do not jump from 6 to 384. Two intermediate steps cost very little and catch almost everything.

Step 3 is the one that gets skipped and the one that saves the money. A plate of identical wells costs one plate and one cell order, and it tells you your position artefacts, your variability and your realistic detection limit before a single compound is committed. Every screen that discovers its edge effect during hit triage skipped step 3.

The re-quote checklist

A scale-up is a re-quote, not a bigger version of the same order. Send this.

Cells

• Total quantity for the whole campaign, computed from total growth area, not well count
• Price per million at every pack size the supplier offers, ask explicitly, because the tiering is steep and rarely volunteered
• Counting convention: total at freeze or viable post-thaw
• Single-lot availability for the whole quantity, reserved, stated on the order confirmation
• If a single lot cannot cover it: how many lots, and can the split be scheduled so the boundary falls between replicates rather than between conditions
• Whether a single-lot assay bank or assay-ready plated format is available, and its price against the frozen-vial route

Plates and consumables

• Exact plate catalogue number, manufacturer, and the manufacturer’s published growth area per well
• Quantity including at least one spare plate for a repeat run
• Lead time in weeks, in writing, and current stock position
• Price per plate at one pack and at campaign volume, and check whether the tiering runs the expected way, because it does not always
• Confirmation of compatibility with your dispenser, imager and reader, by instrument model

Design

• Well count committed to controls per plate, and the control layout
• Edge mitigation: randomisation / perimeter buffer / humidified enclosure, named
• Dispensing mode and the overage factor applied
• Number of plates, and whether they run in one session or across weeks

Schedule

• Delivery date for cells and plates on the same purchase order, with the ship day specified
• Whether the campaign requires a second delivery, and whether that will be from the same lot

Commercial

• Repeat-order price at campaign volume for a second campaign in six months
• Remedy for a lot failing its viability specification, and the claim window
• Whether the supplier will hold reserved material and release it against a schedule

Where we fit

We do this arithmetic as part of normalising quotes, because pack size and counting convention are exactly what make two prices non-comparable, and at screening volume the pack-size question moves the total more than the choice of supplier does.

Send us the plate format, the density, the well count and the replicate structure, and we will come back with the total growth area, the vial count under both manufacturers’ published growth areas, the pack-size comparison across suppliers, and confirmation on single-lot availability and plate stock before you raise a purchase order.