How Much Coverage Do You Need for Long-Read Low-Pass Sequencing?
Coverage is the first number researchers ask about when planning a long-read low-pass sequencing project. How many times does each base need to be read? How does that change for polyploid species? What happens to variant detection when coverage drops?
These are the right questions. The answers are more nuanced than a single number, but they are not complicated once the underlying logic is clear.
This post explains how sequencing coverage works in the low-pass context, what real-world minimum thresholds look like across different use cases, and how to use coverage as a study design variable rather than a fixed assumption.
What Sequencing Coverage Actually Measures
Sequencing depth refers to the average number of times each base pair in the genome has been sequenced. A depth of 2x means that, on average, every position in the genome is represented by 2 reads. Sequencing coverage refers to the breadth of the genome that is sequenced at all — typically expressed as a percentage of bases meeting a minimum depth threshold.
The word 'average' matters. Coverage is not uniform. Some regions will be covered by more reads and some by fewer. Low-depth regions create gaps in variant detection. The goal in study design is to choose a depth that achieves adequate coverage across the fraction of the genome relevant to your research question.
In diploid organisms, coverage is often reported as total coverage or per-haplotype coverage. A diploid genome has two copies of each chromosome. At 4x total coverage, each haplotype is covered at approximately 2x. For long-read low-pass sequencing, per-haplotype coverage is the more useful number because the phasing and variant calling algorithms operate on individual haplotypes.
Why Long-Read Low-Pass Can Operate at Lower Coverage
Depth requirements are not the same across sequencing technologies. Long reads require less depth for equivalent variant detection because of a fundamental difference in how they align to a reference genome.
A short read is 150 base pairs long. In repetitive regions, centromeres, telomeres, and structurally complex loci, reads this short cannot map to a unique location. They get discarded. Effective coverage in these regions drops sharply below the nominal average.
A long read is ~10,000 to 25,000 base pairs. Reads this length span repetitive elements entirely, mapping uniquely to locations that short reads cannot reach. The same nominal coverage depth produces more usable data in complex genomic regions.
At 2x depth per haplotype, a 13-kilobase HiFi read aligns with approximately twice the confidence of a 150-base-pair short read. Genotype recovery for SNPs reaches approximately 85% at 1x and 95% at 4x in human validation experiments — compared to a 30x long-read standard.
Coverage Ranges for Common Research Applications
SNP Genotyping and GWAS
For genome-wide association studies and population structure analysis based primarily on SNPs, coverage in the range of 1x to 2x per haplotype is generally sufficient for long-read low-pass sequencing. At this depth, SNP calls in mappable genomic regions are reliable. Imputation substantially mitigates heterozygous call limitations at the lower end of this range when a phased reference panel is available.
Structural Variant Detection
Structural variant detection in long-read data does not require the same depth accumulation as SNP calling. A single spanning read that crosses both breakpoints of an SV is sufficient to define the variant. At 2x per haplotype, most loci will have at least one spanning read for SVs large enough to be biologically relevant. For rare SV discovery, higher coverage is preferable. For common SV detection across a breeding population, 2x per haplotype is a reasonable working threshold.
Haplotype Phasing
Phasing quality correlates more with read length than with coverage depth for long-read data. A single read spanning a heterozygous SV and flanking SNPs provides direct physical phasing across that segment. Multi-kilobase phase blocks are achievable at low coverage when reads are long enough to span multiple heterozygous sites in a single molecule.
Methylation Analysis
Direct detection of 5-methylcytosine (5mC) from PacBio HiFi kinetics is possible at low coverage. Population-level methylation profiling and differential methylation analysis across haplotypes can be performed with the same sequencing data used for variant calling. Coverage requirements for regional methylation analyses are similar to those for SNP genotyping.
How Genome Size and Ploidy Affect Coverage Decisions
Coverage is a ratio: reads sequenced relative to genome size. A fixed sequencing budget produces different effective coverage depending on how large the target genome is.
For a human genome at approximately 3 gigabases, achieving 2x per haplotype requires a specific number of long reads. For a diploid crop like soybean at 1 gigabase, the same number of reads produces higher per-haplotype coverage. For an allotetraploid like peanut at 2.5 gigabases, the calculation must account for four subgenome copies contributing to effective depth.
Ploidy adds a complication that long-read sequencing handles far better than short-read. In an autotetraploid, reads must be assigned to specific haplotypes for genotyping to be accurate. Short reads are often ambiguously mappable between haplotypes. Long reads span enough heterozygous sites to enable direct haplotype assignment.
For polyploid genomes, effective per-haplotype coverage must account for ploidy level. A tetraploid genome requires twice as many reads to achieve the same per-haplotype depth as a diploid genome of the same total size.
Multiplexing and the Economics of Coverage
Coverage requirements interact directly with multiplexing to determine study economics. Long-read low-pass sequencing supports multiplexing of up to 96 samples per PacBio Revio sequencing cell, depending on genome size and target coverage depth.
A project targeting 2x per haplotype in a medium-sized genome can accommodate more samples per cell than a project targeting 4x per haplotype in a large genome. Coverage decisions are not made in isolation from budget and sample count. The practical design workflow:
Determine the minimum coverage required for your primary research question.
Calculate the reads required to achieve that coverage given your target genome size and ploidy.
Determine how many samples can be multiplexed per cell at that coverage depth.
Compare the cost per sample at that multiplexing level against the per-insight value of the data produced.
The Coverage Tradeoff: What You Gain and What You Trade
More coverage is not always better. It is always more expensive. The relevant question is where the inflection point is for your specific research question.
For SNP-based population genetics and GWAS, data shows diminishing returns above 2x to 4x per haplotype in long-read low-pass runs. You gain confidence in heterozygous calls but are not discovering qualitatively different biological signal.
For structural variant discovery in rare populations, higher coverage adds genuine sensitivity. If rare SVs are your primary target, 4x to 6x per haplotype may be justified.
For most plant breeding applications — building marker panels for selection, characterizing germplasm diversity, identifying loci associated with agronomic traits — the range of 1.5x to 3x per haplotype is where long-read low-pass sequencing provides its best cost-per-insight value.
The Bottom Line
Coverage in long-read low-pass is measured per haplotype. 2x per haplotype is the working standard for most population-level applications.
Long reads require lower depth than short reads for equivalent variant detection because they map more uniquely across complex genomic regions.
SNP genotyping is reliable at 1x to 2x per haplotype for long reads, particularly with imputation.
Structural variant detection does not require deep coverage in long-read data; a single spanning read defines a variant.
Ploidy multiplies coverage requirements proportionally.
Coverage decisions should be made in context of study question, genome size, ploidy, and multiplexing economics — not as a fixed assumption.
Not sure how much coverage your project needs? That question is the starting point for every study design conversation we have.
Talk to a scientist or request a quote.
Related Reading
