Oligo Pool QC Thresholds for CRISPR, DMS and MPRA
Compare published QC threshold bands for pooled plasmid libraries, CRISPR screens, deep mutational scanning, MPRA, and gene assembly. Use this page when you need cited ranges for coverage, barcode support, read depth, and post-correction error burden before interpreting oligo pool quality.
For a shorter interpretation of representation, dropout, Gini, and error-rate metrics, use the Oligo Pool QC guide.
Interpret oligo pool QC targets with published threshold bands instead of broad vendor summaries.
April 22, 2026
Treat these as evidence-backed working bands, not universal regulatory standards or a substitute for assay-specific pilot data.
Use the quality metrics guide for quick interpretation, then confirm critical thresholds against cited sources.
Oligo Pool Guide
Return to the main design, QC, vendor, and ordering guide.
Quality Metrics Guide
Review representation, dropout, Gini, and error-rate interpretation.
Research Library
Review related technical references for oligo pool design and QC.
Synthesis Guide
Review how synthesis choices affect ordering, QC, and rework decisions.
Uniformity Estimator
Estimate representation and dropout risk before finalizing a pool design.
Published QC Threshold Bands
The table below is intentionally conservative. When a paper or protocol publishes a clear floor, it is shown directly. When the literature is more heterogeneous, the band is framed as a practical threshold zone rather than a hard cutoff.
| Use case | Primary QC gate | Working threshold band | Investigate or rework when |
|---|---|---|---|
Any pooled plasmid library before screening Official protocol guidance | Input-library integrity and representation after amplification | Treat a diagnostic digest plus NGS as the minimum QC package. During bacterial amplification, target at least 1000 transformants per perturbation before proceeding. | Re-check the pool if transformant yield falls below the 1000x floor or the digest shows a large recombinant band. |
CRISPR pooled screening Published protocol guidance | Coverage per sgRNA during transduction, selection, and passaging | Use >500 cells per sgRNA as the published floor for standard pooled screens. Move above that floor for weaker or negative-selection designs, with published CRISPRi growth screens often run around 1000x coverage. | Treat sub-500x coverage, strong transduction bottlenecks, or repeated passaging loss as reasons to re-run QC before trusting hit calls. |
Deep mutational scanning and mutagenesis libraries Published protocol guidance | Read depth per variant and base-level subassembly coverage | Published DMS protocol guidance uses about 100 reads per barcode or variant for frequency quantification and at least 100x coverage per base when subassembling variant libraries. | If many variants sit below those depth bands, treat enrichment scores as underpowered rather than biologically meaningful. |
MPRA and barcode-linked reporter libraries Published analysis filters plus QC framework | Barcode complexity per oligo plus DNA/RNA sequencing coverage | A practical published filter is to retain oligos backed by at least 10 barcodes. Exact read depth per oligo depends on library size, but recent large MPRA studies still ran at very high DNA and RNA depth to keep barcode coverage broad. | If a large share of oligos falls below the 10-barcode floor or barcode coverage collapses unevenly between DNA and RNA, treat activity estimates as unstable. |
Gene assembly from oligo pools Published pathway-synthesis and review evidence | Post-correction error burden, not just library completeness | Use less than 1 error per kb as a practical post-correction target. Published pathway-synthesis methods reduced microchip-pool error from roughly 14 errors per kb to about 0.5 errors per kb after correction. | If the corrected pool still sits above the sub-1-per-kb zone, expect clone screening and downstream assembly rework to grow quickly. |
Any pooled plasmid library before screening
This is the baseline gate before downstream application-specific thresholds matter. If the input pool is already skewed, every later readout inherits that bias.
Treat a diagnostic digest plus NGS as the minimum QC package. During bacterial amplification, target at least 1000 transformants per perturbation before proceeding.
CRISPR pooled screening
In pooled CRISPR screens, abundance counts are the phenotype. Under-coverage turns ordinary sampling noise into false dropout or false enrichment.
Use >500 cells per sgRNA as the published floor for standard pooled screens. Move above that floor for weaker or negative-selection designs, with published CRISPRi growth screens often run around 1000x coverage.
Deep mutational scanning and mutagenesis libraries
DMS experiments estimate fitness from frequency changes. If the input library is thinly sampled, the score distribution is dominated by counting error rather than selection.
Published DMS protocol guidance uses about 100 reads per barcode or variant for frequency quantification and at least 100x coverage per base when subassembling variant libraries.
MPRA and barcode-linked reporter libraries
Barcode replication is the noise-control layer in MPRA. Once barcode support gets too thin, plasmid abundance and RNA output become hard to separate from sampling variance.
A practical published filter is to retain oligos backed by at least 10 barcodes. Exact read depth per oligo depends on library size, but recent large MPRA studies still ran at very high DNA and RNA depth to keep barcode coverage broad.
Gene assembly from oligo pools
Assembly projects can tolerate some representation noise, but they fail hard when synthesis error burden stays high across long constructs.
Use less than 1 error per kb as a practical post-correction target. Published pathway-synthesis methods reduced microchip-pool error from roughly 14 errors per kb to about 0.5 errors per kb after correction.
Interpretation Notes
Floors beat averages
For pooled libraries, the most useful thresholds are usually the minimum acceptable floor for coverage or representation, not the mean value after the fact.
Application changes the dominant risk
CRISPR screens are coverage-sensitive, MPRA is barcode-sensitive, and gene assembly is error-burden-sensitive. The same pooled readout should not be judged with one universal cutoff.
Published protocols still need local validation
These bands are meant to anchor decisions, not replace pilot runs. Low-input samples, unusual construct lengths, and custom cloning flows can still justify tighter limits.