siRNA library screens and the perturbation toolkit: knockdown or knockout, and at what scale

Designing a perturbation experiment

1. Choose knockdown or knockout from the biology. Essential genes and dose response questions suit partial, reversible silencing. Clean loss of function suits cutting. Neither is generally better and the choice should be stated with a reason.
2. Solve delivery before scaling. In a screen, delivery efficiency multiplies through every well. Establish it with a fluorescent control and a known positive perturbation before committing library scale material.
3. Size the screen on representation. Library scale work sets a minimum number of cells per reagent that has to be maintained at every step. Calculate it backwards from library size, because a screen below representation is noise.
4. Use an inducible system where timing matters. Where a perturbation is lethal or where the timing of onset is the question, an inducible system separates the perturbation from the selection. It adds a construct and answers a question a constitutive system cannot.
5. Budget the plasmid. Screens consume more DNA than plans allow for. Large scale preparation, or a preparation service, belongs in the plan rather than discovered mid screen.
6. Keep sample preparation enzymes separate in the design. Proteases, glycosidases and nucleases used to prepare samples are analytical tools. Recording them alongside perturbation reagents confuses two different parts of the experiment.

Arithmetic before biology

Library size, cells per reagent, transduction efficiency, selection survival and sequencing depth multiply into a required culture scale that is usually larger than the first plan. That calculation is the screen design.

Do it on paper before ordering anything. Screens fail on representation far more often than on reagent quality.

Controls make a null result usable

A screen with no positive control cannot distinguish a biological negative from a technical failure, and at screen scale a technical failure is expensive.

Include known positive and non targeting reagents distributed across every plate, and check them before analysing anything else.

crispr library screening, and what the pooled format demands

A pooled CRISPR screen is a counting experiment: a library of guides is delivered at low multiplicity, the population is selected, and the guides are read out by sequencing before and after. What that demands is coverage at every step, several hundred cells per guide at infection and at every passage, a selection window long enough to separate the effect from noise, and a readout depth that counts the library rather than sampling it. The library's own quality matters too, the guide distribution at the start and the fraction of the plasmid pool that actually packaged.

crispr cas9 screening libraries, and the shapes they come in

A screening library is a pool of guides delivered as a plasmid pool or as packaged virus, and it comes in three shapes. Genome-wide libraries carry several guides per gene across every gene and need the most cells and sequencing. Sub-libraries target a family, the kinases, the druggable genome, which is what most laboratories should start with. Arrayed libraries put one guide per well, which removes the deconvolution and costs more per gene. The library's own quality, guide representation and the fraction that packaged, is measured before the screen, not after.

A crispr cas9 vector and what it has to carry

A crispr cas9 vector carries the nuclease, a guide cassette, a selection marker and, where stable expression is wanted, an integrating backbone, and packaging efficiency falls as the insert grows. A vector expressing both nuclease and guide is convenient and large; separating them allows a stable nuclease line with a guide library on top, which is what most screens use.

A crispr negative control and what it has to match

A crispr negative control is a non targeting guide delivered at the same multiplicity, selected the same way and carried in parallel, because the delivery, the nuclease and the selection all have effects of their own. A scrambled sequence still has to be checked against the genome, since a control that cuts somewhere is worse than none at all.

Common questions

Silencing or cutting for a screen?

Silencing for essential genes, partial loss and fast, reversible perturbation. Cutting for clean loss of function and for a selection based readout over longer timescales. Both are used and the choice follows the readout.

What makes a screen fail?

Representation, almost always: too few cells per reagent at transduction, selection, sorting or sequencing. The remedy is arithmetic done before the screen rather than analysis afterwards.

Why include a fluorescent delivery control?

Because it separates a failed perturbation from a failed delivery. Without it, a null screen result cannot be interpreted at all.

Browse by service class

• Compare Gene synthesis on advertised prices