Whole-genome sequencing
Inherited variants from blood or saliva DNA, usually at about 30× with a PCR-free library, called with a joint-genotyping pipeline and checked against a truth set.
Our ruleSimulations prepare you for the bench but don't replace hands-on practice.
Germline
Inherited variants from blood or saliva DNA, usually at about 30× with a PCR-free library, called with a joint-genotyping pipeline and checked against a truth set.
How it differs from the standard method
- Heterozygous variants sit near 50% allele fraction, so moderate depth is enough.
- Contamination, sex and identity checks matter because a wrong sample passes every run metric.
- Family (trio) sequencing removes inherited background and finds de novo variants.
1Every base, sampled many times
Whole-genome sequencing breaks a genome into random fragments, reads them, and puts the reads back on a reference (or assembles them when there is none). No region is selected in advance, so coding and non-coding sequence, copy-number changes and structural rearrangements all appear in the same data.
Mean coverage
Bases sequenced ÷ genome size, counted after duplicates and unmapped reads are removed. A 30× human genome needs roughly 110–120 Gb of raw output.
Coverage spread
Even at a good mean, GC bias, mappability and sampling leave some regions thin. The fraction of the genome at ≥ 20× matters as much as the mean.
What a variant looks like
About 50% of reads for a germline heterozygote, 100% for a homozygote. A somatic mutation sits at purity × clonality ÷ 2, often 5–30%.
2Three questions, three designs
Germline sequencing looks for what a person inherited. About 30× is enough to call almost every heterozygous site in the mappable genome.
3How it fails
- Too little output: depth falls below what a heterozygote needs and calls are missed silently, as reference calls.
- Too little input: a library with few distinct molecules gives duplicates, not depth. Sequencing more only returns more copies.
- PCR in the library: GC-rich promoters and first exons are under-covered and homopolymer indels appear. PCR-free libraries avoid both when there is enough DNA.
- Contamination or a sample swap passes every run metric. Only allele-fraction-based contamination estimates and identity SNPs catch it.
- Formalin-fixed tumour DNA carries deaminated cytosines that read as C>T at low allele fraction, mimicking subclonal mutations.
- Without a matched normal, every rare inherited variant looks somatic.
- Short reads cannot assemble across repeats longer than the fragment, so bacterial assemblies stop at rRNA operons and insertion sequences.
Key terms
Practise these as flashcards →Average number of distinct reads covering each base, after duplicates and unmapped reads are removed.
Fraction of reads that are copies of a molecule already read; set by library complexity against sequencing depth.
Share of reads at a position that carry the alternative allele.
Share of true variants that were called, and share of calls that are true, measured against a truth set.
Fraction of cells in a tumour sample that are cancer cells; it scales every somatic allele fraction.
Calls from many unrelated normal samples, used to remove recurrent technical artefacts from somatic call sets.
A contig is a continuous assembled sequence; contig N50 is the length at which half the assembly sits in contigs at least that long.
Phred-scaled accuracy of an assembly: Q40 is one error in 10,000 bases, Q50 one in 100,000.
When to use it
- Rare-disease diagnosis when the causal variant could be anywhere, including outside exons or as a structural or copy-number change.
- Cancer genomes: mutational signatures, structural rearrangements and copy number alongside point mutations.
- Population and family studies that need uniform genome-wide genotypes.
- Complete reference genomes of bacteria and other organisms with no close reference, using long reads.
Compared with alternatives
Far cheaper per sample and much deeper on the genes of interest; prefer it when the question is limited to coding regions or a known gene list.
The platform for repeats, structural variants, phasing and closed assemblies; long-read genomes are increasingly used for germline work too.
The run itself: loading, cluster density and Q30. A genome is only as good as the run behind it.
Where input, fragmentation and PCR cycles are decided, which set duplicates and GC evenness here.
Sub-techniques
Choosing one here switches the whole page, including the protocol and the lab, to it.
Inherited variants from blood or saliva DNA, usually at about 30× with a PCR-free library, called with a joint-genotyping pipeline and checked against a truth set.
- Heterozygous variants sit near 50% allele fraction, so moderate depth is enough.
- Contamination, sex and identity checks matter because a wrong sample passes every run metric.
- Family (trio) sequencing removes inherited background and finds de novo variants.