Skip to main content

Samples and Extraction

Every sequencing run starts with something physical: a tube of blood, a piece of a tumor, a dish of cells. What the sample is and how it was handled decide what the data can show and which errors it carries, long before any instrument is switched on.

Kinds of Samples​

  • Tumor tissue comes from a biopsy or from surgery. It is never pure tumor. Blood vessels, immune cells and connective tissue are mixed in, and the share of cells that are tumor, called the tumor purity, sets how strongly the tumor's variants show up in the data. Low purity weakens the signal of copy number changes, as the copy number variation page shows.
  • A normal sample carries the person's inherited, germline DNA. Blood is the usual source: a standard draw yields plenty, and the white blood cells in it carry the person's full genome. (Mature red blood cells have no nucleus and no DNA.) Saliva or a cheek swab works too and needs no needle. On its own, a normal sample is all that's needed to look for inherited variants. Sequenced alongside a tumor from the same patient, it's called a matched normal. Variants found in both are germline and can be set aside, leaving the somatic ones, as described in A Second Baseline. In leukemias, the blood itself carries the cancer, so the normal has to come from somewhere else, such as a small skin sample.
  • Cell-free DNA floats in blood plasma, released by cells as they die. Its fragments are short, with one study finding a peak at 166 bases (Lo et al., 2010), about the length of DNA wrapped around one nucleosome plus a short linker. In healthy people, most of it comes from dying blood cells (Snyder et al., 2016), but in a person with cancer, some comes from the tumor. That makes it a second way to sample a tumor, with different strengths. Tumor tissue needs a biopsy or surgery and shows one piece of the tumor at one moment. Cell-free DNA needs only a blood draw, which can be repeated to follow a tumor over time, and it can carry DNA from tumors anywhere in the body. The catch is that tumor DNA is often a tiny share of what's there, so finding it takes very deep sequencing. Sequencing cell-free DNA to study a tumor is called a liquid biopsy. The same idea lets prenatal tests read a fetus's DNA from the mother's blood.

Samples also come from cultured cell lines and model organisms such as mice, where the same choices about tissue and handling apply.

Keeping DNA and RNA Intact​

Once tissue is removed from the body, its molecules start to break down, RNA fastest of all. There are two main ways to stop the clock.

Fresh frozen tissue is frozen in liquid nitrogen as soon as possible after collection and stored at very low temperature. It gives the best DNA and RNA, so it's generally preferred when it's available. But it needs a freezer chain from the operating room to the lab.

FFPE (formalin-fixed, paraffin-embedded) tissue is soaked in formalin, which locks molecules in place, then embedded in a block of paraffin wax. FFPE was developed for looking at tissue under a microscope, long before sequencing, and it is still how pathology labs keep nearly every biopsy. The blocks keep for decades at room temperature, so archives hold millions of samples, many with years of follow-up on how each patient did.

The cost is in the molecules. Formalin breaks DNA into short pieces and chemically damages bases. The most common damage turns cytosine into uracil, which the sequencer reads as thymine, so FFPE data carries false C-to-T changes (G-to-A on the other strand) at low fractions that look like real variants (Do and Dobrovic, 2015). Treating the DNA with an enzyme that removes uracil, before it is copied, sharply cuts these false changes (Do et al., 2012). Even so, low-fraction C-to-T calls from FFPE samples deserve suspicion, and important ones are worth confirming with a second method (Do and Dobrovic, 2015).

Extraction and Quality Checks​

Extraction breaks cells open, strips away proteins and other debris, and keeps the DNA or RNA. Before a sample goes further, the lab checks how much there is, how clean it is, and how intact it is.

For DNA, the amount is usually measured with a fluorescent dye that binds only double-stranded DNA. An older method, which measures how much UV light the sample absorbs, also picks up RNA and broken-down DNA, so it reads high, especially for FFPE samples. It is still useful for spotting leftover protein or chemicals from extraction (Simbolo et al., 2013). Intactness is measured by sorting the DNA by fragment size, which gives a DNA Integrity Number (DIN). In one study of FFPE tumor samples, those with a low DIN were far more likely to fail sequencing, though more than half of them still worked (Hiramatsu et al., 2023).

For RNA, the most common integrity measure is the RNA Integrity Number (RIN), which scores a sample from 1, badly degraded, to 10, intact (Schroeder et al., 2006).

These measurements usually travel with the data as sample metadata. When a sample's results look strange, its preservation method and quality scores are the first things to check.