Estimating Sequencing Coverage
`Before starting a sequencing experiment, you should know the depth of sequencing you want to
`achieve. This Technical Note helps you estimate that coverage.
`
`Next-generation shotgun sequencing approaches require sequencing
`every base in a sample several times for two reasons:
`
`• You need multiple observations per base to come to a reliable
`base call.
`
`• Reads are not distributed evenly over an entire genome, simply
`because the reads will sample the genome in a random and
`independent manner1,2. Therefore many bases will be covered
`by fewer reads than the average coverage, while other bases
`will be covered by more reads than average. You need to
`account for this in your planning.
`
`This is expressed by the coverage metric, which is the number of
`times a genome has been sequenced (the depth of sequencing). For
`applications where you aim to sequence only a defined subset of
`an entire genome, like targeted resequencing or RNA sequencing,
`coverage means the amount of times you sequence that subset. For
`example, for targeted resequencing, coverage means the number of
`times the targeted subset of the genome is sequenced.
`
`This Technical Note provides information on how to calculate the
`coverage required for an experiment, and how to estimate the number
`of flow cells or lanes you need to use.
`
`Coverage Requirements Depend on Application
`Illumina does not have an official recommendation for sequencing
`coverage level.
`
`Most users determine the necessary coverage level based on the type
`of study, gene expression level, size of reference genome, published
`literature, and best practices defined by the scientific community. For
`example, the level of coverage for human genome mutations/SNPs/
`rearrangements detection that most publications require is from 10×
`to 30× depth of coverage depending on the application and statistical
`model. For ChIP-Seq studies where reads map to only a subset of a
`genome, often the researchers/publications require coverage around
`100×.
`
`For RNA sequencing, determining coverage is complicated by the fact
`that different transcripts are expressed at different levels. This means
`that more reads will be captured from highly expressed genes, and
`few reads will be captured from genes expressed at low levels. When
`planning RNA sequencing experiments, researchers usually think in
`terms of numbers of millions of reads to be sampled. The number of
`reads required will depend on how sensitive the experiment needs
`to be for genes expressed at low levels. Detecting rarely expressed
`genes might require an increase in the depth of coverage.
`
`Standards are Set by Field and Journals
`The standards are ultimately set by journals and the scientific field you
`are in. The Publications section on Illumina’s website (http://science.
`illumina.com/science/publications/publications-list.html) provides a
`resource for users to search publications for Whole Genome Rese-
`quencing, De Novo Sequencing, Targeted Resequencing, Transcrip-
`tomics and many other fields. This is recommended as a starting
`point for determining the target depth of coverage for a particular
`study. Another good resource for RNA Sequencing is provided by the
`ENCODE project:
`
`http://genome.ucsc.edu/ENCODE/protocols/dataStandards/
`ENCODE_RNAseq_Standards_V1.0.pdf
`
`Estimating Sequencing Runs
`Coverage Equation
`
`The Lander/Waterman equation is a method for computing coverage1.
`The general equation is:
`
`C = LN / G
`
`• C stands for coverage
`
`• G is the haploid genome length
`
`• L is the read length
`
`• N is the number of reads
`
`So, if we take one lane of single read human sequence with v3 chem-
`istry, we get
`
`C = (100 bp)*(189×106)/(3×109 bp) = 6.3
`
`This tells us that each base in the genome will be sequenced between
`six and seven times on average.
`
`Coverage Calculator
`
`Illumina provides an online coverage calculator that calculates the re-
`agents and sequencing runs needed to arrive at the desired coverage
`for your experiment, based on the Lander/Waterman equation. The
`calculator can be found here:
`
`http://www.illumina.com/CoverageCalculator
`
`Perform the following steps to run the calculator:
`
`1. Enter the input parameters:
`
`• The target genome or region size, for example, input 3000 Mb
`(3 Gb) for human genome.
`
`• The coverage you want.
`
`• The total number of cycles. For example, if you want to
`perform 100 bp paired-end runs (2×100), enter 200.
`
`Technical Note: Sequencing
`
`Foresight EX1028
`Foresight v Personalis
`
`

`

`being sequenced a certain number of times. We can use the coverage as
`the average number of occurrences and y as the exact number of times a
`base is sequenced, and then compute the probability that would happen:
`
`P(Y=3) = (6.33 × e-6.3)/3! = 0.077
`Of course, this is the value for exactly 3. It probably is more interesting to
`see the probability the base is sequenced 3 times or less, as most SNP
`callers require at least four calls at a base position to call SNPs. We can
`determine this probability simply by summing up the probabilities for Y=2,
`Y=1, and Y=0:
`
`P(Y<=3) = P(Y=3) + P(Y=2) + P(Y=1) + P(Y=0) =
`
`0.077 + 0.036 + 0.012 + 0.002 = 0.127
`So we see that about 12.7% of the bases in the genome will be covered by
`three or fewer reads, and we will probably want to increase our coverage for
`this experiment.The same formula can be used in a couple other ways. For
`example, by simply computing the Y=0 probability, we can estimate the per-
`centage of a genome not yet sequenced: in our example above 0.2% of the
`genome was not sequenced at all. By multiplying 0.2% by the genome size,
`we see that we would have a total gap length of about 6,000,000 bp. We
`can also estimate the number of gaps by multiplying the number of reads
`used by the percentage of the genome not covered: 0.2% * 189,000,000
`gives 378,000 gaps in the sequence.
`
`2. Select the instruments you want to perform the calculation
`for.
`
`3. Click Submit.
`
`The calculator now writes tables containing the total output required,
`output per lane or flow cell, and number of lanes or flow cells you need
`to use for the desired coverage. You can also download the results
`in a comma-separated values file, so you can share data or use the
`tables in Excel.
`
`Note that the calculator uses an estimate of reads passing filter
`commonly found for balanced genomes (such as PhiX or the human
`genome). If you plan to sequence an unbalanced genome, you may
`have a lower number of reads passing filter, and consequently a lower
`output per lane. If you plan a targeted resequencing or enrichment
`experiment, make sure to read the technical note Optimizing Coverage
`for Targeted Resequencing.
`
`When to Sequence More
`In Illumina sequencing experiments, it is very easy to increase the
`coverage or sequence depth, if you later decide you need more data.
`Provided you still have your original sample, you can just sequence
`more, and combine the sequencing output from different flow cells.
`There are a number of reasons to sequence more than the originally
`estimated coverage, these include:
`
`• The effects you see are not statistically significant. Sequencing
`more reads will generally increase the power of your assay.
`
`• You are investigating events that are very rare. For example,
`you may want to look at transcripts that are expressed at a
`very low level in RNA Sequencing, or look at very low binding
`activities in ChIP Sequencing.
`
`• Certain journals or fields may require a higher level of coverage
`for your particular application.
`
`• Certain genomes may need more sequencing. For example,
`certain regions may be hard to sequence requiring more
`coverage, or the genome may be polyploid.
`
`References
`1. Lander ES, Waterman MS.(1988) Genomic mapping by fingerprinting ran-
`dom clones: a mathematical analysis, Genomics 2(3): 231-239.
`2. Estimating the number of times a base is expected to be sequenced.
`Lander and Waterman made two assumptions about the sequencing:
`• Reads will be distributed randomly across the genome
`• Overlap detection doesn’t vary between reads.
`Based upon these two assumptions, they reached the conclusion that the
`number of times a base is sequenced follows a Poisson distribution. The
`Poisson distribution can be used to model any discrete occurrence given an
`average number of occurrences.The probability function is the following:
`
`P(Y=y) = (Cy × e-C)/y!
`• y is the number of times a base is read
`• C stands for coverage
`We can use the Poisson distribution to compute the probability of a base
`
`Illumina • 1.800.809.4566 toll-free (U.S.) • +1.858.202.4566 tel • techsupport@illumina.com • www.illumina.com
`
`FOR RESEARCH USE ONLY
`
`© 2014 Illumina, Inc. All rights reserved.
`Illumina, other trademarks separated by commas, and the pumpkin orange color are trademarks of Illumina, Inc. and/or its affiliate(s)
`in the U.S. and/or other countries. Pub. No. 770-2011-022 Current as of 12/01/14
`
`Technical Note: Sequencing
`
`Foresight EX1028
`Foresight v Personalis
`
`