sequencing by synthesis
/ SBS /
If you want to read a strand of DNA, one neat idea is to watch it being copied and note down each base as it is added. That is the heart of sequencing by synthesis: you do not read the original strand directly, you build its complement one base at a time and detect, at each step, which base just went in. It is the chemistry behind most next-generation sequencing today.
Millions of identical short DNA fragments are anchored across a chip, each amplified in place into a tiny cluster of copies so its signal is bright enough to see. Then sequencing proceeds in synchronized rounds. In each round, the four bases are supplied as fluorescently labeled, reversibly blocked nucleotides: a polymerase adds exactly one base to every cluster, but the block stops it from adding a second. A camera photographs the whole chip, recording each cluster's color (and thus which base was just added). The block and dye are then chemically removed, and the next round begins. Repeating this for hundreds of cycles, and reading each cluster's sequence of colors, yields a read of that fragment, with millions of fragments read at once across the chip.
Sequencing by synthesis is why genome sequencing got so cheap — its massive parallelism lets one instrument read billions of bases per run. Its trade-offs define its character: reads are short, because errors accumulate as the synchronized rounds slowly fall out of step (a problem called phasing), and homopolymer runs (like AAAAA) can be miscounted. So SBS gives you enormous throughput of short, mostly accurate reads, which is ideal when you can map them to a known reference or read each region many times, but harder when you need to span long, repetitive stretches — that is where long-read methods come in.
On the chip, one cluster flashes green this cycle, then blue, then blue, then red. Decoding the colors gives that fragment's sequence: A-C-C-T, and so on. Millions of other clusters are being read by their own color sequences in the very same photographs.
Add one base, photograph the color, unblock, repeat — across millions of clusters.
Because the synchronized rounds gradually drift out of step (phasing), accuracy falls off toward the end of each read, which is one reason these reads are short — the method buys vast parallel throughput at the cost of read length.