Troubleshoot Low Coverage & Bottlenecks in Pooled Screens

Updated 7 November 2026 · Standard formula, shown below

Pooled mammalian screens (CRISPR, shRNA, ORF) depend on deep, even representation of each guide or barcode. When coverage drops or bottlenecks appear, you risk false negatives and skewed hit lists. This guide walks you through diagnosing the cause—from library amplification to sequencing—and fixing it so your screen stays statistically sound.

What You’ll Need

Your screen’s original plasmid library and the exact guide/barcode list (with counts)
Estimated cell numbers at each step (transductions, selections, harvests)
PCR primers and reagents for library prep (if you’re at that stage)
Access to the sequencing run metrics (Q30, cluster density, read counts per sample)
A spreadsheet or script to calculate fold changes and read distributions
Your protocol’s expected coverage (e.g., 500× or 1000×)

Step-by-Step

Define your expected coverage. Coverage = (number of cells / number of guides) × replicates. For example, a library of 100,000 guides at 500× needs 50 million cells per replicate. Write this number down—you’ll compare every step against it.
Check your starting plasmid library. Sequence the plasmid prep before transduction. If the plasmid library itself has skewed guide representation (e.g., top 10% guides make up 80% of reads), that’s your bottleneck. Re-amplify or re-pool if needed.
Verify transduction efficiency. Use a fluorescent marker or a puromycin kill curve. Aim for 30–50% transduction with a low MOI (0.3–0.5) to ensure one guide per cell. If efficiency is below 20%, you’ll lose a large fraction of guides.
Count cells at every passage. At each split, record the total cell number. If you ever drop below the required coverage (e.g., you need 50M but only have 10M), that’s a hard bottleneck. Thaw more cells or restart.
Monitor selection pressure. For CRISPR screens, ensure the selection (puromycin, blasticidin) is fully killing untransduced cells. If selection is incomplete, non-transduced cells can outgrow and dilute your library.
Harvest enough cells at the endpoint. At the final timepoint, collect at least the same number of cells as your starting coverage (plus a margin for PCR loss). For a 500× screen, harvest 50–100 million cells per replicate.
Extract genomic DNA carefully. Use a column-based kit or phenol-chloroform. Ensure complete lysis and RNA removal. Quantify DNA with a fluorometric assay; low DNA yield means you’re losing coverage.
PCR amplify with enough input. Use at least 1–2 µg of gDNA per PCR reaction. This represents ~150,000–300,000 cells (assuming 6.6 pg per diploid cell). Run enough reactions to cover your entire library (e.g., 100 µg gDNA = 15M cells).
Check PCR cycle number. Keep cycles between 18–22 for CRISPR amplicons. Too many cycles introduce bias; too few give low yield. Run a pilot to find the minimum cycles that give visible product.
Pool and clean up PCR products. Use SPRI beads or gel extraction. Ensure no primer dimers. Run a Bioanalyzer to confirm the correct amplicon size.
Sequence with high depth. Aim for at least 1000× raw reads per guide. For 100,000 guides, that’s 100 million reads per sample. Check your sequencing provider’s output.
Analyze read distribution. After demultiplexing, count reads per guide. Plot the distribution. A healthy screen shows a relatively even distribution with a long tail. If 90% of guides have <10 reads, you have a bottleneck.

Common Mistakes

Underestimating cell numbers. Cells die, stick to tubes, and get lost during media changes. Always start with 2–3× the calculated minimum.
Using too few PCR reactions. One reaction with 2 µg gDNA only covers ~300,000 cells. For a 100M cell library, you need ~300 reactions. Scale accordingly.
Overcycling PCR. Each cycle can amplify bias. Stick to the minimum cycles that yield enough product.
Ignoring sequencing quality. Low Q30 scores cause miscalled bases, which can drop guides out of the analysis.
Mixing replicates. Pooling replicates before sequencing hides bottlenecks. Keep them separate until analysis.

Tips

Always include a plasmid library control sample in your sequencing run. This tells you if the bottleneck is in the library itself or downstream.
Use unique molecular identifiers (UMIs) in your PCR to correct for amplification bias.
Freeze extra cells at each major step. If a bottleneck is detected, you can go back and re-harvest.
Use a computational tool like MAGeCK or casTLE to assess guide distribution and detect outliers.
Run a small pilot screen first to validate your protocol before scaling up.

Troubleshooting

If you see low coverage at the sequencing step: Check whether the bottleneck occurred during cell culture or during library prep. Compare your plasmid library reads to your final reads. If the plasmid library is fine but final reads are skewed, the issue is in transduction or selection—check MOI and kill curves.

If you have high dropout of guides (many with zero reads): This often indicates a bottleneck early on. Look at your cell counts at each passage. Did you ever fall below the required coverage? If so, you may need to repeat the screen with more starting cells.

If PCR yields are low: Increase input gDNA, reduce cycle number, or switch to a high-fidelity polymerase. Check your primer annealing temperature—too high can reduce yield.

If you see uneven representation between replicates: This is normal to some extent, but if one replicate is drastically different, check for batch effects (different media, different virus lots). Consider repeating that replicate.

If your sequencing depth is insufficient: Ask your sequencing facility for more reads, or re-sequence the same library. If that’s not possible, you may need to combine replicates for analysis, but be aware of reduced statistical power.

Frequently Asked Questions

Q: What is the minimum coverage I need for a pooled screen?
A: For most CRISPR screens, 300–500× is the minimum. For shRNA screens, 500–1000× is typical. Higher coverage gives more power to detect small effects.

Q: Can I use frozen cells to recover from a bottleneck?
A: Yes, if you froze cells at an early passage, you can thaw and restart. But you must re-verify that the cells are still in the selection phase and that coverage is adequate.

Q: How many cells do I need for a screen with 100,000 guides?
A: At 500× coverage, you need 50 million cells per replicate at the start. At the endpoint, harvest at least the same number (plus a safety margin).

Q: Why do my guide counts have a long tail?
A: Some guides are inherently less efficient or toxic. A long tail is normal, but if the top 1% of guides have >10% of reads, it suggests a bottleneck or biased amplification.

Q: Can I use a different sequencing platform?
A: Yes, as long as you get enough reads. Illumina is standard, but other platforms work. Just ensure your amplicon length is compatible.

Q: How do I know if my PCR introduced bias?
A: Compare the guide distribution in your PCR product to the original plasmid library. If the distribution changes significantly, reduce PCR cycles or use a different polymerase.

Method: the standard formula for this calculation, rounded as each result states. See how the calculators work or report an error.