Sequencing Depth - How much is Enough?

Advances in DNA sequencing have made genome analysis more accessible than ever. As sequencing costs continue to decline, more organisms are being sequenced. Public sequence databases now contain genomic information from for hundreds of thousands of species, ranging from well-established model organisms to newly sequenced taxa. However, the quality of the resulting data strongly depends on the sequencing technology and, importantly, the sequencing depth.

For example, the budding yeast Saccharomyces cerevisiae has a relatively small genome of approximately 12 Mb, making it inexpensive to sequence at high coverage. Because its genome has been sequenced and annotated extensively, a high-quality reference genome is available, enabling reliable identification of single-nucleotide polymorphisms (SNPs), structural variants, and other genomic features.

In contrast, organisms with exceptionally large genomes pose a much greater challenge. Lungfish are among the most extreme examples, with genome sizes exceeding 40 Gb in several species. Such enormous genomes substantially increase sequencing costs, computational requirements, and the complexity of downstream analyses.

What is sequencing depth?

Sequencing depth, also referred to as coverage, describes how many times, on average, each nucleotide in a genome has been sequenced. For example, a coverage of 20× means that each base has been read approximately twenty times. Greater sequencing depth generally increases confidence in variant detection because true genetic variants can be distinguished more reliably from sequencing errors. However, higher coverage also increases sequencing costs and data storage requirements. Therefore, determining the appropriate sequencing depth is an important part of experimental design.

The required coverage depends on several factors, including:

• the biological question being addressed

• the availability and quality of a reference genome

• the sequencing technology with its inherent error rate,

• the genome size and complexity, (repetitive elements, homopolymers, GC content)

• whether the genomic region of interest is conserved or highly repetitive,

• the type of analysis (e.g., genome assembly, SNP detection, or metagenomics).

What research questions can sequencing answer?

The optimal sequencing strategy depends on the scientific question. For organisms with well-characterized reference genomes, whole-genome sequencing (WGS) is routinely used to identify SNPs, characterize structural variants, perform population genetic analyses, and investigate evolutionary relationships. However, whole-genome sequencing of species with extremely large genomes may not always be the most practical or cost-effective approach. Detecting SNPs or genotyping specific loci can be particularly difficult if the regions of interest are highly repetitive or poorly assembled. In these situations, targeted sequencing offers an attractive alternative. By enriching only selected genomic regions, a much higher coverage of the loci of interest can be achieved while substantially reducing sequencing costs and simplifying downstream analyses.

Targeted versus non-targeted sequencing

Depending on the research objective, sequencing can either focus on the entire genome or only on selected regions.

Whole-genome sequencing (non-targeted) provides an unbiased view of the genome and is suitable for:

• de novo genome assembly,

• comprehensive variant discovery,

• comparative genomics,

• and evolutionary studies.

Targeted sequencing focuses sequencing effort on predefined regions, reducing costs while increasing coverage in those regions. Common approaches include:

• Amplicon sequencing (e.g., 16S rRNA gene sequencing for bacterial community profiling),

• PCR-based target enrichment,

• Hybridization capture using probe panels, most popular example is the Exome sequencing (sequencing only protein coding regions of a genome)

• Restriction enzyme-based methods (e.g., RAD-seq or GBS),

• CRISPR-Cas enrichment, particularly for long-read sequencing of difficult genomic regions.

Choosing between targeted and whole-genome sequencing depends on whether comprehensive genomic information or high-resolution analysis of specific loci is required.

Is there a rule of thumb for sequencing depth?

Although the optimal sequencing depth depends on the organism and research question, the following recommendations provide useful starting points:

Application Typical sequencing depth

Human whole genome sequencing ~30 to 50x (genotyping to causative variants)

Human whole exome sequencing ~50 to 200x (genotyping to somatic variants)

Bacterial whole-genome sequencing ~100× coverage

Hybrid bacterial assembly (ONT + Illumina) ~100× ONT and ~20–30× Illumina

De novo assembly of larger genomes Often >50× long-read coverage

16S rRNA amplicon sequencing Approximately 50,000–80,000 reads per sample

Shotgun metagenomics Approximately 5–10 Gb of sequencing data per sample (depending on community complexity)

Calculating coverage

For whole-genome sequencing, coverage can be estimated using the formula:

Coverage = (Read length × Number of reads) / Genome size

For example, sequencing a 5 Mb bacterial genome with 2 million paired-end reads of 150 bp yields 2,000,000 × 2 × 150 bp = 600 Mb of sequence, or approximately 120× coverage.

A note on transcriptomics

For RNA-based applications, "depth" means library size rather than coverage per base. Reads are distributed by transcript abundance, not genome size, so the requirement is driven by the dynamic range of the transcriptome. Three points follow from this:

• Replicates beat depth. Beyond ~10–20 million reads per sample, adding biological replicates raises statistical power more than adding reads. The exception is isoform and splicing analysis, where 50–100 million paired-end reads are appropriate.

• Only usable reads count. rRNA carry-over, adapter dimers and duplicates reduce the raw output.

• Metatranscriptomics spans 15–150 million because rRNA depletion efficiency and community evenness vary so widely.