In 1998, Shankar Balasubramanian and David Klenerman from Cambridge University's Chemistry Department developed a novel DNA sequencing method. Their approach would soon become the most popular sequencing technology in the world.

The main idea of their method is simple; during DNA replication, we can add nucleotides with different fluorescent labels, detect the light emitted, and determine which nucleotide was attached.

This workflow is described as follows. (Figures in this FAQ are taken from an excellent primer on sequencing technologies, Mardis 2008. You may wish to consult her paper for additional details.)

• The DNA is chopped up into smaller fragments that are modified by adding adapters to each end (figure below, left) - short, synthetic fragments of DNA.

One end of each modified DNA fragment is anchored on the surface of a chip called a flow cell (figure below, right). The flow cell is covered by primers, which are short sequences that are complementary to the adapters

The instrument cannot detect signal from a single molecule, so we need to obtain multiple copies of each DNA fragment, located in a particular place on the flow cell surface. The free ends of adapters that have been anchored will bend and hybridize to nearby complementary primers. If we just add nucleotides and DNA polymerase, DNA replication will begin, extending from the primer around the arch to the adapter on the opposite end of the modified DNA fragment. The arching double strands of DNA look like bridges, so this process is called bridge amplification (figure below, left).

In a denaturation step, we cleave the two strands of DNA joined in each bridge. As a result, we have twice the amount of single-stranded DNA that we began with (see figure below, right).

After the first round of bridge amplification we have two copies of the molecule, attached nearby. We then repeat this process several times; after each round, the number of copies doubles, resulting in a large number of identical DNA fragments located near each other and forming clusters. This process occurs across the whole flow cell, and so we obtain a large number of these clusters on the cell surface, as shown below.

We can then cut one end of the bridges using a restriction enzyme, and all the reverse strands will be washed off the flow cell, leaving only forward strands attached. This process occurs across the whole flow cell.

Then we start sequencing itself. Primers are added, and fluorescently tagged nucleotides called reversible terminators are added to the cell to base pair with single strands. Each of the four bases emits its own color when a laser is fired at the DNA, so we take a photo of the flow cell, which tells us which nucleotide is being incorporated at each spot in each cycle. Following the incorporation step, an enzyme cleaves off the fluorescent dye, unused nucleotides and DNA polymerase molecules are washed away, and nucleotides are added in order to base pair with the next nucleotide down on each strand and begin the imaging process again. You can see an animation of this process at the following video.

With each cycle of sequencing, there is signal decay and molecules in the cluster that fall behind or jump ahead in the synthesis process, leading to higher noise. This explains why we cannot read long DNA fragments.

For a video illustrating this process, please watch the following video produced by Illumina.

Yes, but the resulting overlap graph may potentially have more edges. In practice, edges in an overlap graph are formed by pairs of reads that overlap significantly but not perfectly (in order to account for errors in the reads).

No. In practice, we won't actually glue the nodes; instead, we will simply create a node for each distinct prefix and suffix that appear in our k-mers.