PacBio DNA Sequencing: SMRT and HiFi Reads

PacBio DNA sequencing is single-molecule, long-read sequencing: a polymerase copies one DNA molecule inside a tiny well on a SMRT Cell and the instrument records the incorporation of each base as flashes of light. For most researchers the practical question is not how the chemistry works but what the sequencing core hands back and how to turn it into an alignment. This guide maps the read vocabulary (ZMW read, subread, CCS, HiFi read) onto the files you receive, then walks through aligning HiFi reads with pbmm2, the aligner PacBio maintains.

What it is used for in research

PacBio describes its HiFi sequencing as sequencing native DNA without PCR, and lists these applications on its technology page (checked 2026-10-29):

Single-nucleotide variants, indels, structural variants and tandem repeat expansions.
5mC methylation calls with phasing, taken from changes in polymerase kinetics rather than bisulfite conversion.
Full-length isoform RNA sequencing (Iso-Seq) and single-cell applications.
Targeted sequencing by hybrid capture or PCR.
Complete genome assembly, including telomeres and centromeres. For choosing assembly software, see choosing a de novo genome assembler and how N50 is calculated and why it is not enough alone.

How it works, in PacBio’s own description: sequencing happens in millions of tiny wells on a SMRT Cell; a circular SMRTbell template lets the polymerase read the same molecule repeatedly; and the repeated passes are combined by Circular Consensus Sequencing (CCS) into one HiFi read. If you are comparing platforms rather than processing PacBio data, see Illumina vs Nanopore for the short-read and nanopore side.

From the instrument to your files: the read vocabulary

The terms below come from the PacBio file-format documentation (version 13.0, checked 2026-10-29) and explain what a core facility can deliver.

Term	Meaning in the PacBio documentation
SMRT Cell	A silicon chip on which SMRT sequencing runs.
ZMW	Zero-mode waveguide: one well on the SMRT Cell, identified by a hole number.
ZMW read	The full basecaller output for one well.
Subread	The span of a ZMW read that corresponds to the DNA insert.
CCS read	A consensus built from the subreads of one molecule. The CCS tool takes an unaligned subreads BAM as input.
HiFi read	A CCS read with predicted accuracy of QV 20 or higher.

The practical consequence: if you receive an unaligned BAM of HiFi reads, those are already consensus reads and you do not run CCS yourself. If you receive subreads, CCS is the step that produces HiFi reads. Ask your core which of the two you are getting. A Phred quality of 20 means an expected error rate of 1 in 100; for the FASTQ quality encoding, see FASTQ format explained.

Reading the vendor accuracy and length claims

PacBio’s pages do not state read length and accuracy in a single form. Compare the wordings before you quote a number in a methods section, and attribute the figure to PacBio.

PacBio page (checked 2026-10-29)	What it says
HiFi sequencing technology page, comparison table	Accuracy 99.95% (Q33); the same table lists short-read sequencing-by-synthesis at 99.92% (Q31) and nanopore at 99.26% (Q21).
HiFi sequencing technology page, body text	Accuracy given as 99.9%; read length up to 25 kb, and a FAQ range of 500 to 20,000 bases.
SMRT sequencing page	“Over 99% accuracy” for individual HiFi reads; reads of tens of kilobases, with 15 to 25 kb typical in the FAQ.

These are vendor statements and the figures differ by page, by chemistry and by what they measure. Cite the page and the date you read it, and report the quality values from your own run rather than the marketing figure.

Install or access (official source only)

The sequencing run itself is normally done by an instrument owner or a core facility. PacBio’s technology page names Revio as the production instrument and Vega as a benchtop instrument. For a typical native whole-genome library, PacBio lists 500 ng of input (SPRQ chemistry), roughly 20x coverage as typically sufficient for variant and methylation calls, and a typical human HiFi BAM of 30 to 60 GB. Check the input requirements with your core before you send DNA.

For the analysis step, pbmm2 is published by Pacific Biosciences on GitHub (github.com/PacificBiosciences/pbmm2) under the BSD-3-Clause-Clear license. Its README says the latest version can be installed via the bioconda package pbmm2, with setup details in the PacBio bioconda repository (github.com/PacificBiosciences/pbbioconda). Install the bioconda package of that name into a dedicated environment.

Step-by-step task: align HiFi reads to a reference with pbmm2

Confirm what you hold. The input can be BAM (aligned or unaligned), FASTA or FASTQ (optionally gzipped), or a .Q20.fastq file. A HiFi BAM or FASTQ is what you want.
Prepare the reference as FASTA, a .referenceset.xml, or a prebuilt .mmi index.
Run the maintainers’ own example, replacing the file names and read-group values with yours:
```
pbmm2 align ref.fasta movie.Q20.fastq ref.movie.bam --sort --rg '@RG\tID:myid\tSM:mysample'
```
Leave the preset alone for HiFi data. The README lists the presets SUBREAD, CCS or HIFI (the default), ISOSEQ and UNROLLED, and says presets vary by version, so run pbmm2 align --help for your installed version.
Take the output as a BAM. If you omit the output file name, the BAM is streamed to standard output.

Output and how to read it

The result is an aligned BAM; the PacBio documentation notes that alignment tools take unaligned BAM files and produce aligned BAM. The --sort option in the example above sorts the output, and the --rg string sets the read group, including the sample name in SM. For how BAM compares with SAM and CRAM, see SAM vs BAM vs CRAM. To look at the alignments, IGV is one viewer. Because HiFi reads are defined as QV 20 or higher, expect read-level quality at or above that threshold; for run-level quality, use the numbers your core reports.

Troubleshooting

Unexpected preset or option error. The pbmm2 README states that presets vary by version. Run pbmm2 align --help on the installed version instead of copying a command from an older protocol.
Subreads instead of HiFi reads. A subreads BAM is the input to CCS, not a finished HiFi set. The documentation defines HiFi reads as CCS reads with predicted accuracy of QV 20 or higher, so generate or request those first.
Output printed to the terminal. pbmm2 streams BAM to standard output when no output file is named; name the output file.
Using minimap2 with the wrong PacBio preset. minimap2’s README separates map-pb (PacBio CLR reads) from map-hifi (PacBio HiFi/CCS reads, version 2.19 and later) and says the difference matters for the read type.

Free alternatives (table)

Tool	What its own documentation says
pbmm2 (Pacific Biosciences)	Default preset is CCS or HIFI; accepts BAM, FASTA and FASTQ; BSD-3-Clause-Clear license; installable via bioconda.
minimap2 (Heng Li)	Lists `-ax map-hifi ref.fa pacbio-ccs.fq.gz > aln.sam` for PacBio HiFi/CCS genomic reads (v2.19+), `-ax map-pb` for CLR reads, and `-ax splice:hq -uf` for Iso-Seq. Output is SAM; a bioconda install is linked from its README.

Whichever you choose, use one aligner per project and record its version.

FAQ

What is PacBio HiFi sequencing?

It is PacBio’s name for long reads built by Circular Consensus Sequencing: a polymerase reads a circular SMRTbell template repeatedly and the passes are combined into one consensus read, called a HiFi read when its predicted accuracy is QV 20 or higher.

Does PacBio sequencing need PCR or bisulfite treatment?

PacBio states that HiFi sequencing reads native DNA without PCR and detects 5mC methylation from polymerase kinetics, so bisulfite conversion is not needed for modified bases.

What is the difference between subreads and HiFi reads?

Subreads are the spans of one well’s read that correspond to the DNA insert; a HiFi read is the consensus of those subreads for a molecule, with predicted accuracy of QV 20 or higher.