Arrayed versus pooled screens

A 96-well plate and a single transduced flask can both be entered in a notebook as "the screen". They answer different questions and fail in different ways. Arrayed versus pooled screens is the choice between knowing the reagent from the well coordinate and knowing the reagent from a sequence counted later. The nuclease those reagents carry is described in how CRISPR-Cas9 editing works in research. Confirming an individual edit, once a hit has a name, is in checking whether a genome edit worked.

One reagent, one well

An arrayed screen puts a single guide, or another single perturbation, in each well. The phenotype is read in that well: a viability dye, an image, a reporter, a supernatant. Identity is the plate map. If well C4 was annotated as a particular spacer, the result in C4 belongs to that annotation until you show the map was wrong.

The cost of that clarity is scale and handling. Genome-wide arrays consume plates, tips, and time. Edge wells evaporate differently from centre wells. A missed dispense is a missing reagent, and it looks like a negative phenotype if you do not have a way to see that the well was empty of guide. Liquid-handler logs and a dye or volume check are part of the experiment.

Arrayed layouts make controls local. Non-targeting wells on the same plate see the same medium and the same incubation. A positive-control guide that should kill or should activate a reporter should do so on that plate. If it does not, the plate's transfection or the nuclease batch is the first suspect, not the fifty experimental guides. How those controls are chosen more generally is in controls for a CRISPR experiment.

What an array does not give you is a genome-wide ranking from one culture. It also does not prove the cut. A phenotype in one well, with one guide, is compatible with an off-target. A second independent guide in another well, moving the same phenotype, is the usual arrayed answer to that doubt.

A library in one population

A pooled screen delivers many guides together, often as a lentiviral library, into one population of cells. Each cell should, by design, receive one guide. The culture is then selected or challenged. At the end, and at the start, the guide cassettes or their barcodes are amplified and counted by sequencing. Guides that become more abundant are enriched. Guides that become rarer have dropped out.

Representation is the number of cells, or the number of reads, that stand behind each guide at the beginning. A guide present in only a few cells can vanish from sampling noise. Library protocols therefore ask you to keep a stated excess of cells per guide and to sequence deeply enough to see the library. Those numbers come from the library's own design notes. Copying a figure from a different library is how representation collapses while the viability looks fine.

Multiplicity of infection is the average delivery events per cell. Low multiplicity reduces the chance that two guides occupy one cell and confuse which spacer caused the phenotype. High multiplicity increases that confusion and can also increase the fraction of cells that must be removed because they were not transduced as you intended. Measure transduction in your cells. Do not assume a titre measured in another line.

Sequencing is not optional decoration. Without it, the flask has no well coordinates. Read counting follows the same general short-read ideas as in next-generation sequencing from library to reads. A PDF of the top ten gene names, without counts, a sample sheet, or the starting library, is not a screen result.

Enrichment, dropout, and essential genes

Under a drug or another selection, guides that help cells survive become a larger share of the reads. That enrichment is a hypothesis about resistance, still compatible with an off-target of a single guide. Several guides against one gene, moving together, are stronger than one spacer.

Dropout is the mirror. Guides against essential genes should be lost as edited cells die or stop dividing. Screening groups use that pattern as a control even when the biological question is elsewhere. If the essential-gene guides do not drop, cutting may have failed, representation may have been too thin to see loss, or the culture time may have been too short for a growth difference to appear. If they drop and your candidate does not, you have a negative result worth keeping, not a failed gel.

Non-targeting guides should stay near their starting abundance, within the noise of counting. A non-targeting set that drifts in one direction suggests a bottle effect, a PCR bias, or a selection you did not mean to apply. Positive controls should move in the direction the screen was built to detect. Both behaviours belong in the methods, next to the candidate list.

Where each format answers the wrong question

Use an array when the phenotype cannot be reduced to abundance. A change in cell shape, a localisation, or a timed image is hard to recover from a barcode count unless you add a sorting or single-cell step on purpose. Use a pool when the question is growth, drug survival, or another trait that changes how many cells carry that guide, and when you can keep representation.

A pool used to chase a microscope phenotype, with no sort, will not see it. An array used to claim a genome-wide essential-gene map, from three plates, has not covered the genome. A fluorescent reporter of the vector shows delivery in either format and still does not show a cut.

PCR of the cassette can distort counts. Too many cycles, or primers that prefer some guides, invent enrichment. Sequence the starting plasmid library or the day-zero cells and compare. A guide missing at day zero cannot drop out later. It was never there.

Culture conditions that move the counts

In a warm laboratory, a pooled flask left on a bench during a delay keeps growing, and growth is the assay. Representation shifts before the drug is added. Log the time out of the incubator. An arrayed plate with a loose lid loses volume from the edges and concentrates the drug. Those edge wells need a place in the analysis, or a layout that does not put every control in the centre and every hit on the rim.

Power cuts stop both the incubator and, later, the sequencer. A screen that sat at the wrong temperature is not rescued by a deeper sequencing run. Note the interruption and decide, with the controls, whether the culture is still the experiment you designed.

Viral libraries remain a biosafety question for the institution that holds the cells. The choice of array versus pool does not lower containment. Research screens are not clinical tests. A hit list is not a therapy.

What to specify when reagents or reads are the enquiry

State arrayed or pooled, the cell line, the library or the guide count, the delivery class, and whether you need plasticware, arrayed RNAs, or a discussion of counting reads. Put that on the quote request. Classes of enzyme and nucleic acid sit in the molecular biology catalogue. If a hit must be confirmed as an on-target allele, the CRISPR validation sequencing enquiry reference is a discussion prompt and an independent method reference. Ask whether a quotation is possible. The molecular biology pathway is the research setting for the assay you will actually read.