How to Decide If Oligo Pool Error Rate and Coverage Are Good Enough

Use this page when you need to decide whether an oligo pool is ready to move forward. It shows how to turn vendor error-rate claims into full-length risk, QC read-depth targets, and clear proceed-vs-rescue-vs-re-order decisions. If you need a faster answer on one step, jump to our Error Rate Calculator, Coverage Calculator, QC Metrics Guide, and Troubleshooting Checklist.

Judging oligo pool error rate and sequencing coverage before release

Convert read depth and vendor claims into a real go or no-go decision for your pool.

Key Takeaways

  • •Convert any published per-base error rate into per-oligo risk before you decide a pool is acceptable. Longer oligos accumulate errors quickly even when the headline rate sounds strong.
  • •Coupling efficiency determines full-length product yield: at 99.5% per-step efficiency, a 100-mer yields only ~61% full-length product. At 99.0%, this drops to ~37%.
  • •Per-member coverage (λ=3.0 for 95%) is not the same as whole-library coverage. For a 10,000-member pool to reach 95% probability that every member is represented, you need λ ≈ 10.6, or about 106,000 total reads.
  • •For QC and uniformity review, plan far above presence/absence minimums. A 10K pool typically needs 5-10 million reads to judge dropout, skew, and representation with confidence.
  • •Published vendor pages do not all expose the same QC detail. Compare the readout format and ask for per-sequence data when a public page omits numeric uniformity or error metrics.
  • •Rescue small dropout lists individually when the rest of the pool is sound. Re-order when red flags cluster in critical targets, representation is poor, or correction would erase too much yield.

What are you trying to decide?

→ Is my vendor error rate acceptable for this oligo length?→ How much full-length product should I expect?→ How many reads do I need for QC?→ Can I rescue a weak pool instead of re-ordering?→ Which published vendor specs actually matter?→ Should I proceed, rescue missing oligos, or re-order?

Related QC checks

→ Read Oligo Pool QC Metrics→ Calculate per-oligo error risk→ Calculate QC read depth→ Estimate uniformity before ordering→ Compare vendor QC packages→ Troubleshoot dropout, skew, and cloning failures
Page contents

Which Synthesis Errors Actually Matter?

Oligonucleotide synthesis (primarily via phosphoramidite chemistry) introduces three categories of errors. Understanding each type is essential for predicting their impact on downstream applications.

Error TypeMechanismFrequencyImpact
DeletionIncomplete coupling + capping failure~1 in 200–1,000 ntFrameshifts in coding sequences; most common error
SubstitutionImpure phosphoramidites, depurination~1 in 500–3,000 ntMissense mutations in proteins; may be tolerable
InsertionDetritylation failure, double coupling~1 in 1,000–5,000 ntFrameshifts; least frequent error type

Error rates vary significantly by synthesis platform. Traditional column synthesis typically has higher error rates (~1:200 nt) than modern array-based platforms (~1:2,000–3,000 nt). See Section 2 for details. Reference: LeProust, E.M. et al. (2010). "Synthesis of high-quality libraries of long (150mer) oligonucleotides by a novel depurination controlled process." Nucleic Acids Research, 38(8): 2522–2540.

What Error Rate Should You Expect from Each Platform?

The synthesis platform changes both the baseline error rate and the way you should interpret it. Use platform-level numbers as a starting point, then convert them into per-oligo risk for your actual sequence length.

PlatformTechnologyError RateMax Length
Column synthesisTraditional phosphoramidite~1:100–1:200 nt~200 nt
Microarray (Agilent)Inkjet printing on glass~1:1,500–1:2,000 nt230 nt
Silicon chip (Twist)Silicon-based massively parallel~1:2,000–1:3,000 nt350 nt
Electrochemical (CustomArray)Electrochemical deprotection~1:500–1:1,000 nt170 nt

Interpretation: Error rates should be evaluated in context. For a 100-nt oligo at 1:2,000 error rate, the expected number of errors per oligo is 100/2,000 = 0.05, meaning ~95% of oligos will be error-free. For a 300-nt oligo at the same rate, 300/2,000 = 0.15, so ~86% will be error-free. Longer oligos inherently have more total errors.

Error-rate comparison note: When comparing vendor error rates, always ask whether they report "per-base error rate" (errors per nucleotide synthesized) or "per-oligo error rate" (fraction of oligos with any error). A 1:2,000 per-base rate for 100-mers means ~5% of oligos have errors — this sounds very different from "95% accuracy," but they're the same number.

How Much Full-Length Product Should You Expect?

In stepwise oligonucleotide synthesis, each coupling step has a finite efficiency. The fraction of full-length product decreases exponentially with sequence length:

Full-length yield = (coupling efficiency)N−1

where N is the sequence length in nucleotides. This means that even small differences in coupling efficiency have dramatic effects on yield for long oligos.

Coupling Efficiency50-mer Yield100-mer Yield150-mer Yield200-mer Yield
99.5%78.0%60.7%47.3%36.8%
99.0%61.3%37.0%22.4%13.5%
98.5%48.0%22.6%10.6%5.0%
98.0%37.5%13.8%5.1%1.9%

Verification: 99.5% coupling, 100-mer: 0.99599 = 0.607 (60.7%) ✓  |  99.0% coupling, 100-mer: 0.99099 = 0.370 (37.0%) ✓

How Many Reads Do You Need for Oligo Pool QC?

When determining how many sequencing reads are needed for an oligo pool, it is critical to distinguish between per-member and whole-library coverage. Confusing these two leads to dramatically insufficient sequencing depth.

Per-Member Coverage (Single Oligo)

The number of reads for any single member follows a Poisson distribution with mean λ (average reads per member). The probability that a specific member has at least one read:

P(X ≥ 1) = 1 − e−λ
λ (avg reads/member)P(≥1 read) per memberReads for 10K pool
1.063.2%10,000
2.086.5%20,000
3.095.0%30,000
5.099.3%50,000
10.099.995%100,000

Whole-Library Coverage (All Members)

Coverage interpretation check: Many protocols state "λ=3 gives 95% coverage" without clarifying this is the per-member probability. For a library of N members, the probability that every member has ≥1 read is:

P(all members covered) = (1 − e−λ)N

For a 10,000-member library at λ=3: P = (0.950)^10,000 ≈ 5.3 × 10−223 — essentially zero! The 5% chance of missing any single member compounds across 10,000 members.

Library Size (N)λ for 95% whole-libraryTotal Reads Needed
100~6.6~660
1,000~8.9~8,900
10,000~10.6~106,000
100,000~12.9~1,290,000

Derivation

To find λ for 95% whole-library coverage of N members: solve (1 − e−λ)N ≥ 0.95. Taking logarithms: N × ln(1 − e−λ) ≥ ln(0.95). This gives: 1 − e−λ ≥ 0.951/N, so e−λ ≤ 1 − 0.951/N, and λ ≥ −ln(1 − 0.951/N). For N=10,000: λ ≥ −ln(1 − 0.951/10000) = −ln(1 − 0.999994872) ≈ 10.6.

Practical recommendation: For QC validation of an oligo pool, target ≥100× average coverage (λ=100). This provides enough reads to assess uniformity, detect dropout members, and estimate error rates from the sequencing data itself. For screening experiments (e.g., CRISPR), follow library- specific coverage guidelines from the screening protocol.

Coverage check: Don't confuse per-member coverage (λ=3 gives 95%) with whole-library coverage. For a 10K library, λ=3 means each individual oligo has a 95% detection chance — but the probability that all 10,000 are detected is near zero. You need λ≈10.6 (106K reads) for 95% whole-library coverage.

Error-rate interpretation check: Vendor-reported error rates ("1 in 1,800 nt") measure per-base accuracy after correction. But your functional experiment cares about full-length correct fraction. For a 200 nt oligo at 1:1,800 error rate, the probability of any error is 1 − (1799/1800)²⁰⁰ ≈ 10.5%. That means ~90% correct — fine for most pooled screens, but problematic for clonal work where you need >99% accuracy. Always convert per-base rates to per-oligo rates using our Error Rate Calculator.

Planning QC Reads for a 10K Library

You've received a 10,000-oligo CRISPR library from Twist Bioscience. How many sequencing reads do you need for QC? Let's work through the math.

Define the Coverage Goal

Goal: 95% probability that every oligo in the pool has ≥1 read (whole-library coverage).

Calculate Required Lambda

Using the whole-library formula: λ ≥ −ln(1 − 0.951/10,000) = −ln(1 − 0.999994872) ≈ 10.6

Calculate Total Reads

Total reads = λ × library_size = 10.6 × 10,000 = 106,000 minimum reads

Add Margin for Non-Uniformity

Real pools aren't perfectly uniform. Twist reports >90% of oligos within 2× of mean, meaning some oligos have 0.5× expected reads. To compensate:Target 500–1,000× average coverage = 5–10 million reads. This ensures even underrepresented oligos (×0.5) get 250–500 reads.

Result: Sequencing plan

For a 10K pool: allocate 5M reads on a MiSeq nano run (≈500× coverage). Cost: ~$200–400 depending on read length. This gives you quantitative uniformity data, not just presence/absence.

Coverage calculation note: Use our Coverage Calculator to compute reads needed for your specific pool size and confidence level. It handles both per-member and whole-library calculations automatically.

What Can You Do If Error Rates Are Too High?

Post-synthesis error correction can reduce error rates by 3–10×, improving the fraction of functional sequences in your pool. Two main approaches are available:

Enzymatic Mismatch Cleavage

Principle: Denature and re-anneal the oligo pool so that error-containing strands hybridize with correct strands, forming mismatched duplexes. Treat with a mismatch-specific endonuclease that cleaves at or near the mismatch site, then select for full-length surviving strands.

Common enzymes:

  • T7 Endonuclease I — recognizes and cleaves heteroduplex DNA at mismatches, insertions, and deletions
  • CEL nuclease (Surveyor) — similar activity to T7 EndoI, often used in mutation detection
  • Commercial kits: ErrASE (Novici Biotech), CorrectASE — optimized multi-enzyme procedures

Performance: Typically reduces error rates 3–10× but also reduces total yield by 30–70%. Best suited for applications requiring high-fidelity sequences (protein variant libraries, gene-length constructs).

Sequence Verification by NGS + Selection

Principle: Clone the oligo pool into a sequencing vector, sequence individual clones by NGS, and computationally select only error-free sequences for downstream use. This is the most thorough approach but requires cloning and is expensive for large pools.

When to use: Small, high-value libraries (<1,000 members) where error-free sequences are critical (e.g., saturation mutagenesis, defined variant libraries for structural studies).

Error-correction note: Enzymatic error correction works best when the majority of oligos are correct (i.e., error rate <1:500). For pools with very high error rates (>1:200), the correction enzyme will cleave so many strands that yield becomes impractically low. In that case, re-order from a better vendor rather than trying to "fix" a bad synthesis.

Protocol: T7 Endonuclease I Error Correction (20 μL)▾
Step 1 — Denature/Re-anneal the oligo pool
─────────────────────────────────────────────
Oligo pool (100 ng/μL) 10 μL
10× NEBuffer 2 2 μL
Nuclease-free H₂O 8 μL
─────────────────────────────────────────────
95°C 5 min → ramp -2°C/sec to 85°C →
ramp -0.1°C/sec to 25°C (hold)
Step 2 — T7 EndoI digestion
─────────────────────────────────────────────
Re-annealed pool 20 μL
T7 Endonuclease I (10 U) 1 μL
─────────────────────────────────────────────
37°C for 15 min (short oligos) or
37°C for 30 min (gene-length constructs)
Step 3 — Inactivate & purify
─────────────────────────────────────────────
65°C 20 min → column cleanup (Zymo)
Expected yield: 30-70% of input
Expected error reduction: 3-10×

For long oligo pools, pilot digestion conditions before scaling the correction procedure. Over-digestion reduces yield without improving error rate. Always run a pilot with 10% of pool first. Source: NEB T7 Endonuclease I Protocol; Hughes et al., Nat Methods 2017.

Which Published Vendor Specs Should You Compare?

When you compare vendors, separate published ordering specs from the performance questions you still need confirmed by the vendor. The table below keeps to fields published on current official product pages.

SpecificationTwist BioscienceIDT oPoolsAgilent SurePrint / HiFi
Published oligo lengthUp to 350 nt40-350 bases30-230 nt across surfaced catalog SKUs
Published pool size / scaleNo limits on pool size1 pmol: 100-20,000 oligos; 10 pmol: 10-2,000; 50 pmol: 2-384Catalog SKUs surfaced up to 100,000 unique sequences depending on length
Published QC detail>90% within <2.0x mean; error rate up to 1:3000Current oPools page markets high fidelity and low dropout, but lists QC and quantification as noneSurePrint emphasizes uniformity and monitoring; HiFi advertises error rates as low as 1 in 4000 nt
Published turnaround2-4 business days4-7 business days from order to deliveryStandard: as few as 3 business days; HiFi: as few as 5 business days
Platform languageSilicon chipProprietary synthesis platform used for oPoolsMicrofluidic DNA writing; HiFi tier for higher-fidelity pools

What to ask before approving a quote: representation target, dropout threshold, whether the vendor provides raw per-sequence read counts, whether pool-level sequencing is included, and whether the published turnaround includes only production or full delivery.

Vendor specification details are aligned to the current OligoPool vendor comparison and public vendor specification review. Specifications can change by product tier, sequence length, and quote structure, so always confirm current terms before ordering.

Should You Proceed, Rescue Missing Oligos, or Re-Order?

You've received your pool and run NGS QC. Use this decision matrix to decide whether the pool is fit for purpose, whether a targeted rescue is enough, or whether the failure is large enough to justify a re-order.

QC MetricProceedProceed with cautionRe-order or correct
Library representation≥95% oligos detected85–95% detected<85% detected
Uniformity (Gini)Gini ≤0.2Gini 0.2–0.4Gini>0.4
Fold-change rangeMax/min <5×Max/min 5–10×Max/min >10×
Error rate≤1:1,500 nt1:500–1:1,500 nt>1:500 nt
Dropout clusteringRandom distribution<3 dropouts per pathway>3 dropouts in critical pathway
Full-length fraction≥80%60–80%<60%
Correct insertion (%)≥90% (for cloned libs)70–90%<70%

When to proceed

  • All metrics are green or at most one is yellow
  • No dropouts in your "must-have" gene set
  • Adequate replicates to compensate for noise

When to re-order

  • Any metric is red AND affects your biological question
  • Essential gene controls are underrepresented
  • Dropout clusters in a critical pathway

Rescue option note: Before re-ordering, consider spike-in rescue: order the missing oligos individually (~$5–10 each) and add them to your existing pool. This is faster and cheaper than full re-synthesis for <500 dropouts. Use our Error Rate Calculator to model whether your pool quality is sufficient for your screen's statistical power.

Frequently Asked Questions

How many sequencing reads do I need for my oligo pool?▾
This depends on whether you need per-member coverage or whole-library coverage. For per-member: λ=3.0 gives 95% probability that any single member has ≥1 read (30,000 reads for a 10,000-member pool). For whole-library: to ensure 95% probability that ALL 10,000 members have ≥1 read, you need λ≈10.6 (~106,000 reads). For quantitative uniformity analysis, we recommend ≥500× average coverage (5 million reads for 10K members). Use the formula: reads = λ × library_size.
What is the difference between substitution and deletion errors?▾
Substitutions are wrong-base errors where the correct nucleotide is replaced by another (e.g., A→G). They occur due to impure phosphoramidite reagents or side reactions during coupling. Deletions are missing-base errors where a nucleotide is skipped entirely, usually due to incomplete coupling followed by capping failure. Deletions are typically more frequent than substitutions in phosphoramidite synthesis and more problematic because they cause frameshifts in protein-coding sequences.
How does coupling efficiency affect my experiment?▾
Coupling efficiency compounds exponentially with oligo length. At 99.5% efficiency, a 50-mer yields ~78% full-length product, but a 200-mer yields only ~37%. The non-full-length products contain deletion errors and may produce truncated or frameshifted proteins in expression experiments. For long oligos (>150 nt), consider gene synthesis instead of oligo pool synthesis, or use enzymatic error correction post-synthesis.
How should I compare published vendor QC specs?▾
Compare vendors on four things: published length range, pool-size limits, turnaround, and whether they publish pool-level QC metrics or only general quality language. Twist publishes numeric uniformity and error-rate claims on its Oligo Pools page; IDT oPools publishes pool size, length, and turnaround but lists QC and quantification as none on the surfaced product page; Agilent publishes turnaround for SurePrint and HiFi tiers plus a HiFi fidelity claim. If a metric is not published, ask for representation, dropout, uniformity, and raw per-sequence data before you approve a quote.
What is enzymatic error correction and when should I use it?▾
Enzymatic error correction uses mismatch-cleaving enzymes (e.g., T7 endonuclease I, CEL nuclease, or commercial kits like ErrASE) to selectively degrade oligos containing errors. The process involves: (1) denaturing and re-annealing the oligo pool so error-containing strands hybridize with correct strands; (2) treating with mismatch-cleaving enzyme; (3) amplifying the surviving correct fragments. This can reduce error rates 3–10× but also reduces total yield. Use it when error-free sequences are critical (e.g., protein variant libraries, CRISPR screens).
What does ">90% of oligos within 2× of mean" uniformity mean?▾
This is a common uniformity metric for oligo pools. It means: measure the read count for each oligo in the pool by NGS, calculate the mean read count, then check that ≥90% of all oligos have a read count between (mean/2) and (mean×2). A pool with poor uniformity might have 30% of members over- or under-represented by >2×, leading to uneven coverage in screening experiments. Twist publishes >90% within <2× on its Oligo Pools page; if your vendor does not publish a numeric target, ask for the actual distribution or raw per-sequence counts.

Related Tools

Calculate Per-Oligo Error Risk

Convert per-base error rates and coupling efficiency into functional pool risk.

Calculate QC Read Depth

Size your QC run for per-member detection, whole-library coverage, and experiment scale.

Estimate Uniformity Before Ordering

Model how pool size and read depth affect representation before the vendor report exists.

Read Oligo Pool QC Metrics

Interpret representation, dropout, Gini, fold-range, and error rate in vendor or NGS QC reports.

Fix Oligo Pool QC Problems

Review recovery options for dropout, skew, weak representation, and cloning failures.

Compare Vendor QC Packages

See how vendors differ on QC scope, turnaround, and fit for different applications.

Related Error Rate and Coverage Pages

Continue with the QC, vendor, or troubleshooting page that matches whether you are still planning a pool or diagnosing one that already exists.

Plan Oligo Pool Synthesis from Design to QC

Open synthesis planning if you still need to define platform, quote, or delivery strategy.

What QC Should I Request for Oligos?

Open this when the next decision is choosing vendor QC expectations and acceptance checks.

Read Oligo Pool QC Metrics

Use the metrics page when you need help interpreting representation, dropout, or skew in a report.

Troubleshoot Pool Dropout and Cloning Failures

Move here when the library has already been synthesized and the next job is diagnosis or recovery.

Compare Oligo Pool Vendors

Review vendor fit when QC scope, turnaround, or scale limitations are shaping the decision.