Library prep and exome capture
A sequencer cannot read a whole chromosome, only short DNA fragments with adapters on both ends. Library prep turns genomic DNA into a collection of such fragments; exome capture then picks out the coding regions, so that the exome, about 1–2% of the genome, gets most of the sequencing. Insert size, duplicate rate, on-target rate and capture size are all fixed in these two steps, and the capture size later becomes the denominator of TMB.
Library prep: turning DNA into molecules the sequencer can read
First the DNA is fragmented, acoustically or enzymatically, into pieces a few hundred base pairs long; WES usually aims for a main peak around 150–250 bp. End repair then blunts the ragged ends, and a single A is added at the 3′ end so the fragment can ligate to adapters carrying a T overhang. The adapters hold the sequencing primer sites and a sample index, so many samples can be pooled on one run and separated afterwards by index. Finally PCR amplifies the adapter-ligated molecules to a usable amount.
PCR copies each original molecule several times, and copies that are read more than once are PCR duplicates. The less DNA goes in and the more damaged it is, the fewer distinct original molecules there are. That is low library complexity, and it means more duplicates for the same amount of sequencing.
Hybrid capture: fishing exons out of the whole genome
The finished library covers the whole genome. A capture kit supplies a large set of biotin-labelled probes (RNA or DNA) complementary to the exons. The library is denatured to single strands and hybridised to the probes, then streptavidin-coated magnetic beads are added: biotin binds streptavidin tightly, a magnet holds the probe-bound fragments, and unbound off-target fragments are washed away. The fragments eluted from the beads get one more round of PCR and become the exome library that goes on the sequencer.
Capture is not perfect. Some fragments bind a probe only partly, others are carried down non-specifically, and both become off-target reads. Exons with very high or very low GC content hybridise and amplify less efficiently, so their coverage is noticeably thinner.
Capture size: the denominator of TMB
Every kit ships with a target BED file listing the regions it is designed to capture; their total length is the capture size. Common exome kits cover about 30–60 Mb, the differences coming from probe design, whether UTRs are included and how much padding surrounds each target. The same tumour sequenced with different kits yields different numbers of countable mutations, which is why TMB divides by the capture size (chapter 8). A stricter denominator is the territory actually callable: the part covered well enough in both tumour and normal.
Common pitfalls
- Assuming every exon is covered evenly. GC-rich first exons and regions with close pseudogenes are often thin, so a gene with “no mutation” may simply not have been seen.
- Using the designed territory instead of the callable one. Poorly covered regions cannot yield mutations but still sit in the denominator, which underestimates TMB.
- Sequencing deeper to rescue a low-complexity library. The extra reads are mostly duplicates and effective depth barely moves; more input or a new library is the fix.