Sequencing depth is the average number of times each base in a DNA or RNA sample gets read during a sequencing experiment. If you sequence a human genome at 30x depth, that means every position in the genome was read an average of 30 times. Higher depth gives you more confidence that what you’re seeing is a real biological signal rather than a random error, but it also costs more and generates more data to process.
How Sequencing Depth Is Calculated
The basic formula is straightforward: multiply the total number of reads by the length of each read, then divide by the size of the genome (or target region). If a sequencing run produces 300 million reads that are each 150 bases long, and you’re sequencing a human genome of about 3 billion bases, your depth is (300,000,000 × 150) / 3,000,000,000 = 15x. That “x” notation is standard shorthand, so 30x means each base was covered 30 times on average.
The key word there is “average.” Depth is never perfectly uniform across a genome. Some regions get sequenced far more than 30 times while others might only be covered a handful of times. This unevenness matters because a region with only 2x or 3x coverage is much more likely to contain missed variants or errors than one covered at 30x.
Depth vs. Breadth of Coverage
These two terms sound similar but describe different things. Depth (sometimes called vertical coverage) is how many times a given position was read. Breadth (horizontal coverage) is the percentage of the target region that was sequenced at least once, or at least to some minimum threshold. You can have excellent depth on average but still have gaps where certain stretches of the genome weren’t captured well at all.
This tradeoff shows up clearly when comparing targeted panels to whole-transcriptome sequencing in cancer testing. Targeted panels that focus on a few hundred genes can achieve depth over 500x, while broader approaches covering more than 20,000 genes typically reach only 30 to 50x. The deeper, narrower approach is more sensitive for detecting rare events in those specific genes, while the broader approach can find unexpected findings across the genome.
Why Depth Matters for Accuracy
Every sequencing read has a small chance of containing an error. When you only read a position a few times, it’s hard to tell whether an unusual base is a true variant or just a mistake. Stacking up many reads at the same position lets you distinguish real signals from noise through sheer repetition.
Empirical testing with ultra-deep whole-genome data shows that single-nucleotide variant calling reaches over 95% concordance with a validated truth set at roughly 17.6x depth. For simpler genetic variation, around 15x depth strikes a reasonable balance between accuracy and cost. But detecting small insertions and deletions is harder: even at 13.7x depth, concordance for these variants reaches only about 60%, which is why higher depth is generally needed when indels are clinically important.
For heterozygous variants (where only one copy of a gene carries the change), you need enough reads to reliably see both versions. At very low depth, you might only capture reads from one copy and miss the variant entirely.
Recommended Depth for Different Applications
There’s no single “right” depth. The target depends entirely on what you’re trying to find and how rare or subtle it might be.
- Whole genome sequencing (germline): Clinical labs typically run at 30x to 60x. At 65x depth, one study found that 100% of coding regions in medically important genes recommended by the American College of Medical Genetics were fully covered at 13x or higher, something whole exome sequencing couldn’t match even at 154x average depth due to capture gaps.
- Whole exome sequencing: Usually performed at 100x to 150x. Because exome capture kits only pull out about 1-2% of the genome, you need higher depth on those targeted regions to compensate for uneven capture efficiency. Even so, roughly 2% of coding regions may have incomplete coverage due to inherent limitations of the capture process, particularly in GC-rich stretches.
- Cancer sequencing (somatic mutations): Large-scale cancer projects like The Cancer Genome Atlas use 100 to 150x for exomes and 30 to 60x for whole genomes. At 30x depth, sensitivity for detecting a mutation present in 20% of reads is about 95.6%. But for rare mutations found in only 5% of reads (common in mixed tumor samples), sensitivity at 30x drops to just 16%, improving to about 52% at 60x. At 150x exome depth, mutations at 3% allele fraction can be detected with roughly 66% sensitivity.
- RNA sequencing: Depth is measured differently here, usually as total reads per sample rather than fold-coverage. For standard differential gene expression analysis, 20 to 30 million reads per sample is generally sufficient.
- Long-read sequencing: Platforms from Oxford Nanopore typically need 30 to 50x coverage for genome assembly, while PacBio HiFi reads, which are more accurate per read, can work well at 15 to 25x.
What Causes Uneven Depth
Even when you aim for 60x, not every base in the genome reaches that number. Several factors create peaks and valleys in coverage across the genome.
GC content is one of the biggest culprits. Regions that are unusually rich in G and C bases tend to be harder to capture and amplify, often ending up with lower coverage. Library preparation methods also introduce bias: the specific capture kit, the type of fragmentation, and whether PCR amplification was used all shape how evenly reads distribute across the genome. PCR-free library preparation generally produces more uniform coverage.
A multi-center study analyzing whole exome data from the 1000 Genomes Project found that different sequencing facilities produced significantly different depth distributions, even when analyzing the same set of clinically relevant variants. Each center used different capture kits and protocols, and these methodological differences created systematic biases in which regions were well-covered and which were not. This means that two exome datasets with the same average depth can have meaningfully different diagnostic sensitivity depending on where and how they were generated.
How Depth Affects Experimental Design
Choosing a target depth always involves a tradeoff with the number of samples you can afford to run. Sequencing one sample at 100x costs roughly the same as sequencing three or four samples at 30x. For population studies or differential expression experiments, running more samples at moderate depth usually provides more statistical power than running fewer samples very deeply.
In RNA sequencing, a formal sample size calculation involves five interacting factors: sequencing depth and the expected read count for a given gene, biological variability within each group, the size of the expression change you want to detect, your acceptable false positive and false negative rates, and the number of samples per group. Depth alone can’t compensate for too few biological replicates, because biological variability between individuals typically dwarfs the technical noise that deeper sequencing would reduce.
For cancer genomics, the math shifts because the goal is often detecting very rare mutations rather than comparing groups. Here, pushing depth higher on a single sample directly increases your ability to find low-frequency variants that would otherwise be invisible. The choice between 30x and 150x can mean the difference between detecting a mutation carried by 10% of tumor cells or missing it entirely.