Multiplexing & Coverage in Reduced Representation Genome Sequencing: Depth Targets, Missingness, and Power

You have a fixed budget and a large cohort. You need stable PCA/ADMIXTURE, robust selection scans, or even association-like signals—but you’re unsure how many reads per sample to buy, how far you can multiplex, and how much missingness your reviewers will tolerate. This guide turns those anxieties into a planning workflow you can defend in audits and reviews.

By the end, you’ll have: (1) a depth and multiplex planning workflow; (2) an Input → Depth Range table you can use for scoping and quotes; (3) a framework to judge when missingness threatens publishability; and (4) a copy-paste checklist of inputs we need to quote accurately.

TL;DR — Fast Answers

• If your goal is PCA or ADMIXTURE, prioritize balanced pooling and batch control over chasing extreme depth. See the method-specific budget model in the ddRAD project design resource: ddRAD project design.
• If your goal is selection scans, prioritize loci count and consistent missingness filters.
• If your goal is association-like analysis, consider low-pass WGS when missingness is likely structured. For route comparisons, see SNP arrays vs low-pass vs deep WGS.
• Multiplexing is too aggressive when pool imbalance and locus dropout push missingness beyond your downstream tolerance.
• Always plan a safety margin for pooling variance and sample-quality spread.
• Use the Input → Depth Range table below to convert project inputs into a realistic reads-per-sample range.

What Coverage Means in Reduced Representation Genome Sequencing

In reduced representation sequencing (RRS), coverage is not uniform genome depth—it is the combination of per-locus depth, loci yield, and the missingness pattern. The enzyme pair, size-selection window, and pooling balance together shape this coverage. Foundational work on RAD/GBS shows that ddRAD achieves repeatable locus sets with tunable yield, while GBS pushes multiplexing higher with sparser per-locus depth.

Three common sources of missingness:

• Restriction site polymorphism that causes allele or locus dropout.
• Library and PCR bias that disproportionately depletes certain fragments.
• Pooling and sequence allocation imbalance that leaves part of the cohort under-covered.

Method foundations: ddRAD established dual-digest and size-selection principles for repeatable locus sets in population genomics, as shown in Peterson et al.’s Double Digest RADseq (2012, PLOS ONE). GBS popularized single-enzyme, barcode-heavy workflows enabling very high multiplexing but relying more on imputation, per Elshire et al.’s genotyping-by-sequencing paper (2011, PLOS ONE).

If you’re seeing batch issues or structured missingness, review cohort-scale QC and sampling-bias controls here: QC metrics and batch effects and Sampling bias and batch effects in population genomics.

Start with Your Analysis Goal

Depth targets are not universal; your downstream goal determines how sensitive you are to missingness and genotyping uncertainty.

Goal → what to fear most → what to control first:

• Population structure (PCA or ADMIXTURE): fear structured missingness and batch artifacts → control pool balance and batch QC.
• Diversity and relatedness: moderate sensitivity → control a stable SNP set and consistent filters.
• Selection scans: fear insufficient loci or biased missingness patterns → control loci yield and uniform thresholds.
• Association-like: fear unstable genotype calls and confounding → strongly consider switching to low-pass WGS with genotype likelihoods or SNP arrays. See comparisons: ANGSD vs ddRAD for low coverage WGS and SNP arrays vs low-pass vs deep WGS.

Figure 2. A decision workflow linking downstream objectives to depth targeting, multiplexing with safety margin, and pre-defined QC gates for reduced representation sequencing (RRS) studies.

Use This Input → Depth Range Table to Plan Reads per Sample

If you can describe your samples and your goal, you can translate that into a defensible reads-per-sample range and a realistic multiplexing plan.

For method-specific budget models and in-silico design aids, see: ddRAD project design

Assumptions

• Platform/read length: Illumina PE150.
• Read counting scope: pass-filter read pairs per sample (R1+R2).
• Safety margin: 25% to absorb pooling variance and DNA quality spread.

Input fields

• Organism and genome complexity flag: Small <1 Gb; Medium 1–3 Gb; Large >3 Gb.
• Sample type and DNA quality tier: High (DIN ≥8 or dominant >10–20 kb fragments; clean ratios); Medium (DIN 5–8; some degradation; stricter pooling); Low (DIN <5; consider stratified libraries or alternative routes).
• Cohort size and group structure: per condition and per batch.
• Primary downstream goal: structure, selection, or association-like.
• Risk flags: expected batch heterogeneity, low input, degraded DNA.

Output columns

• Recommended reads per sample (range): Tier 1 = 2–5M; Tier 2 = 5–10M; Tier 3 = 10–20M read pairs.
• Recommended multiplex range (range): compute from lane yield; do not hard-code fixed samples per lane.
• Expected missingness tier: qualitative band tied to goal and DNA tier.
• Must-have QC checks: demultiplex success/index purity; reads-per-sample CV and 0.5×–2× coverage proportion; overall and stratified missingness; effective loci/SNP scale; duplication/over-amplification.

Automatic adjustment rules

• For each drop in DNA quality tier, raise reads per sample by 30–50% or reduce multiplex one level.
• If multi-batch harmonization is critical, add 20–30% reads or tighten QC gates.

Multiplex formula

• Max multiplex ≈ (Lane usable read pairs) / (Target read pairs per sample × (1 + safety margin)), with safety margin default = 25%.

Worked example

• Suppose a lane yields 600M usable read pairs, your target is 6M read pairs per sample (Tier 2 mid), and you keep a 25% safety margin. Max multiplex ≈ 600M / (6M × 1.25) ≈ 600M / 7.5M ≈ 80 samples per lane. If your pilot shows pool CV trending toward 0.30 or <70–80% samples within 0.5×–2× the median, cut this target by 10–20% or stratify libraries.

Set Multiplexing with a 6-Step Checklist

Multiplexing should be set by minimum acceptable depth after pooling variance, not by the maximum number of samples you can technically barcode.

1. Define your minimum acceptable callability for the goal
   • Declare overall and per-group call-rate targets (e.g., ≥90% overall; ≥85% per batch for structure).
   • Write them into your “Assumptions & Gates” page so they are auditable.
2. Estimate loci yield conservatively
   • Use prior runs or in-silico tools to estimate loci/SNP counts for your enzyme pair and size window; if unknown, assume lower yield to keep plans safe. ddRAD enzyme/window predictions are discussed in Lajmi et al.’s ddgRADer paper (2023).
   • Calibrate with a small pilot before scaling.
3. Choose an initial reads-per-sample range
   • Start from the Tier ranges above and adjust for DNA quality and cohort risks.
   • Keep alternatives in mind if your goal pushes the limits of RRS (e.g., association-like).
4. Convert reads-per-sample to multiplex range, then add a safety margin
   • Apply the multiplex formula using realistic lane yields; add the 25% margin by default.
   • Favor unique dual indexes and conservative pooling to limit index bleed and allocation variance.
5. Stress-test against batch and sample-quality spread
   • Simulate or use historical CVs; if CV >0.30 or <70–80% samples fall within 0.5×–2× the median reads in pilots, reduce multiplex or stratify libraries.
   • If you’re planning multi-batch cohorts, review these first: QC metrics and batch effects and Sampling bias and batch effects in population genomics.
6. Lock QC gates and rerun triggers before sequencing
   • Define demultiplex success/index purity, reads-per-sample CV thresholds, missingness targets (overall and stratified), effective SNP counts, and duplication limits. Pre-authorize reruns when any trigger is tripped.
   • If you’re planning multi-batch cohorts, review these first: QC metrics and batch effects and Sampling bias and batch effects in population genomics.

Understand How Missingness Changes Power and What Reviewers Question

High missingness can be survivable; structured missingness (by batch, population, or plate) is what undermines conclusions and triggers reviewer skepticism.

Figure 3. Conceptual relationship between missingness and downstream stability, highlighting the disproportionate impact of missingness that is structured by batch or population.

• Scenario 1 — Random missingness high: If overall missingness is high but random, PCA/ADMIXTURE often remain stable when you control pooling balance and apply consistent filters. Expect wider confidence intervals and more SNPs filtered out, but structure can still be interpretable, especially with ≥20k–50k usable SNPs and pool CV ≤0.25. Report both pre- and post-filter call rates and include sensitivity analyses to reassure reviewers.
• Scenario 2 — Structured missingness moderate: When missingness correlates with batch or population, PCA/ADMIXTURE can split by batch rather than biology. Early warnings include PC1 aligning with run or plate, per-batch call-rate gaps >5–10pp, and consistent low-depth tails within a group. Actions: tighten batch-harmonized filters, reduce multiplex, stratify libraries, or consider switching strategy to low-pass WGS with genotype likelihoods or arrays: ANGSD vs ddRAD and SNP arrays vs low-pass vs deep WGS.
• Scenario 3 — Missingness low but loci too few: Even with low missingness, selection scans lose power if the locus count is small. Below ~50k SNPs, outlier-based or window statistics become unstable; aim for ≥100k for more robust windows, with per-locus depth medians ≥8–12×. If enzyme/window combinations cannot deliver, re-design or add reads before scaling; pilots pay for themselves here.

Control the Budget Drivers That Actually Move the Needle

The cheapest way to protect publishability is to control pool balance, QC gates, and rerun policy—not to guess a higher multiplex and hope.

Controllable

• Lane allocation, pooling strategy, QC thresholds, whether to pilot, and rerun triggers.

Hard to control

• DNA quality distribution, restriction site polymorphism, and batch-source heterogeneity.

Figure 4. Key controllable levers versus hard-to-control sources of variability that drive publishability risk in cohort-scale RRS projects.

For budget models and design best practices, including pilot-first then freeze, see: ddRAD project design.

What We Need from You to Quote Accurately

If you provide the inputs below, we can recommend a depth and multiplexing range with clear assumptions and QC gates you can put into a statement of work.

Copy-paste and fill:

• Species and estimated genome size (give a range if unknown)
• Sample count and grouping
• Sample type and DNA amount/quality notes
• Target analyses (pick one to two primary)
• Whether batches are expected (collection sites, timepoints, plates)
• Any hard constraints (budget ceiling, timeline)

Next steps: scope your project on the ddRAD-Seq service page: ddRAD-Seq service

FAQs

Quality Gates and Templates You Can Reuse

Go or No-Go decision thresholds by goal

Batch QC table template

Summary metrics to compute per run:

• Reads-per-sample CV.
• % samples within 0.5×–2× median reads.
• Structured missingness gap (pp) between groups/batches.

Version freeze and change log template

CD Genomics is mentioned in this guide only where it provides practical context. For example, when scoping a ddRAD cohort, our planning workflow uses your inputs (species, DNA quality tier, cohort structure, target analyses) to propose a tiered reads-per-sample range, estimate multiplexing with a 25% safety margin, and define audit-friendly QC gates and rerun triggers. This example is illustrative; the same planning logic can be implemented in-house with equivalent SOPs and tooling.

  Related Services

Genome-wide Association Analysis Service

Whole Genome Re-sequencing for Population Genetics

Reduced-Representation Sequencing for Population Genetics

Pan Genome Sequencing Service

Pan‑Genome & Adaptive Breeding

Accessory Genome Characterization

Core Genome Analysis Overview

Sequencing

  Download

Power Your Regulatory Discoveries with DAP-seq: A MYB-GLK Case Study

RNA-seq, ATAC-seq, and CRISPR Screening: Powering CAR T-Cell Research

2b-RAD Sequencing: Your Gateway to High-Density SNP Genotyping

Plasmid/Phage Sequencing: Unlocking the Power of Engineering Tools