Sequencing flow cells: from library to base call
Inside the glass slide where millions of DNA fragments are amplified, imaged and read. A plain explainer on how next-generation sequencing (NGS) flow cells turn an extracted sample into the data behind a hereditary-cancer panel — and why clinical labs still confirm with Sanger.
What a flow cell actually is
A flow cell is a small glass slide etched with channels and coated with a lawn of short anchored DNA oligos. Prepared sample fragments — each tagged with adapter sequences — flow across this surface, bind, and are copied in place until each starting molecule becomes a dense, clonal cluster of roughly a thousand identical copies. The sequencer then reads all of those millions of clusters at once, one base per cycle, which is what makes "next-generation" sequencing massively parallel rather than one-fragment-at-a-time. The richer the cluster density and the more cycles run, the more total data — but for a clinical panel, what matters is not raw output so much as read depth: how many independent reads cover each base of each gene. This page walks the path from library to call, and explains why a flagged variant is often re-checked by an older, slower method.
From library to base call
Four stages turn purified DNA from a swab or blood draw into variant calls on a panel such as one covering <code>BRCA1</code>, <code>BRCA2</code>, <code>PALB2</code> and <code>TP53</code>.
- 01
1 · Library preparation
Extracted DNA is fragmented, end-repaired and ligated to sequencing adapters — the universal handles that let fragments bind the flow cell. Sample-specific index barcodes are added so many patients can share one run and be demultiplexed afterward. The library is amplified, quantified and QC'd before loading.
- 02
2 · Cluster generation
The library flows onto the flow cell, where fragments anneal to the surface oligos. Bridge amplification folds each fragment over to copy it repeatedly in place, building millions of discrete clonal clusters — each one a bright, readable signal of ~1,000 identical strands.
- 03
3 · Sequencing by synthesis
Fluorescently labelled, reversible-terminator nucleotides are added one cycle at a time. After each incorporation the flow cell is imaged, the colour recorded per cluster, the terminator and dye cleaved, and the next base added. Paired-end runs read both ends of each fragment for accuracy.
- 04
4 · Alignment & variant calling
Reads are demultiplexed by barcode, aligned to the reference genome, and a caller flags positions that differ from reference. Per-base coverage is checked against the lab's depth threshold; low-coverage regions are noted for orthogonal follow-up before any variant is classified.
Throughput tiers across common platforms
Flow cells come in capacity tiers. Small, fast instruments suit focused panels; large ones suit exomes and genomes. Figures are approximate, vendor-published ranges and vary by read length and run mode.
| Tier / platform | Flow cell | Approx. output | Approx. reads | Typical use |
|---|---|---|---|---|
| Benchtop (MiSeq) | High-output kit | Up to ~15 Gb | ~50M paired-end (2×300 bp) | Small targeted panels, amplicons |
| Mid (NextSeq) | High-output (e.g. P4) | Tens–hundreds of Gb | Up to ~1.8B reads | Larger panels, small exomes |
| High (NovaSeq 6000) | SP / S1 / S2 / S4 | ~80 Gb to ~3 Tb (single) | ~800M to ~10B | Exomes, genomes, large cohorts |
| Ultra-high (NovaSeq X Plus) | 1.5B / 10B / 25B | Up to several Tb | ~1.5B to ~25B read-pairs | Population-scale genomics |
Table 1. Representative Illumina flow-cell output tiers. Output and read counts are manufacturer specifications and depend on read length, kit version and single- vs dual-flow-cell mode.
Read depth by assay type
"Depth" is reads per base. Targeted panels sequence a small region very deeply so even low-level or mosaic variants are caught; whole genomes spread the same machine output thinly across all 3 billion bases.
~30x across the genome
~100x across coding regions
≥100x floor per base
>300x, often ~500x
Glossary
A few terms that recur when reading a sequencing report.
- Cluster
- A clonal spot of ~1,000 identical DNA copies generated on the flow cell by bridge amplification; the unit the imager reads.
- Read depth (coverage)
- How many independent reads span a given base. Clinical hereditary-cancer panels commonly target >300x mean with a ≥100x per-base floor.
- Paired-end
- Reading both ends of each fragment, which improves alignment accuracy and helps detect insertions, deletions and rearrangements.
- Allele fraction
- The proportion of reads at a position carrying the variant. Very low fractions (e.g. <30%) more often warrant orthogonal confirmation.
- Orthogonal confirmation
- Re-checking an NGS call with an independent method — classically Sanger sequencing — to guard against platform-specific artefacts.
“Massively parallel sequencing answers 'what is here?' across millions of fragments at once. Sanger, reading one segment cleanly to ~1,000 bases, answers 'are we sure?' for the calls that change care.”
Why Sanger still confirms
Despite NGS dominating discovery, many clinical laboratories still confirm reportable findings with Sanger sequencing, the original chain-termination method. Sanger produces high-accuracy reads of roughly 800–1,000 base pairs and serves as an orthogonal check: a second, chemically independent look that catches platform-specific artefacts an NGS run might share across reads. Practice is evolving — interlaboratory work shows that high-confidence calls above a defined quality and allele-fraction threshold (for example, allele fractions well above 30%) can be reliably reported without routine Sanger, while lower-allele-fraction or low-coverage calls still benefit from confirmation. Regions an NGS panel covers poorly are also commonly back-filled by Sanger. The throughline: the flow cell finds the candidate; an orthogonal method confirms the ones that matter. None of this is diagnostic on its own — interpretation belongs with a clinician or genetic counsellor.
Common questions
Does a bigger flow cell mean a better cancer-panel result?
Not directly. A larger flow cell produces more total data, but a clinical panel's quality depends on read depth at each targeted base, not raw output. A small benchtop run focused on a few dozen genes can deliver deeper, more reliable per-base coverage than a genome run spread thinly across 3 billion bases.
What read depth does a hereditary-cancer panel need?
Clinical panels commonly target a mean of >300x — often around 500x — with a minimum per-base floor near 100x. Higher depth improves confidence for low-level or mosaic variants and helps distinguish true calls from noise. Exact thresholds are set during each lab's validation.
Why are some variants re-sequenced with Sanger?
Sanger is an orthogonal, chemically independent method. Confirming an NGS call by Sanger guards against platform-specific artefacts, and is especially valued for low-allele-fraction or low-coverage calls and for regions the panel covers poorly. Many labs now reserve it for those harder cases rather than every variant.
What's the difference between coverage and accuracy?
Coverage (depth) is how many reads see a base; accuracy is how often each read is correct. Deep coverage lets the caller average over many reads to overcome individual errors, which is why clinical panels sequence so deeply. Confirmation methods then address accuracy a different way — by reading the segment cleanly end to end.
- [1]NCBI GTR. Genetic Testing Registry — comprehensive hereditary cancer panels (assay coverage and read-depth specifications, e.g. >300x mean, ≥100x minimum).↗
- [2]Illumina. NovaSeq 6000 and NovaSeq X Plus sequencing system specifications (flow-cell tiers, output and read counts).↗
- [3]Thermo Fisher. Next-Generation Sequencing Illumina workflow — library prep, cluster generation, sequencing by synthesis.↗
- [4]CD Genomics. Sanger sequencing for validation of next-generation sequencing — orthogonal confirmation and read-length characteristics.↗
- [5]bioRxiv 2018. A rigorous interlaboratory examination of the need to confirm NGS-detected variants with an orthogonal method in clinical genetic testing.↗
Trace the whole path, sample to sequencer
This is one stop on the lab workflow. See how a sample is extracted before it reaches the flow cell, and how the resulting calls are read on a report — explainers only, not medical advice.