Scatter millions of DNA fragments across a glass slide. Copy each fragment into its own dense little cluster, so that a single invisible molecule becomes a spot bright enough to see. Then add one letter at a time, and arrange the chemistry so that every letter flashes a colour as it goes on.
Photograph the slide. Add the next letter. Photograph it again.
Do that a few hundred times and you have millions of sequences, read at once, from a stack of pictures of a piece of glass.
The idea was worked out around the turn of the century by a small company in Cambridge, England, called Solexa. Illumina agreed to buy them in November 2006 and closed the deal that January, for roughly 600 million dollars in stock.
The chemistry was ingenious. The idea underneath it was almost embarrassingly plain.
Sanger's method read one fragment of DNA per reaction. Careful, accurate, slow, and reading a human genome that way took the Human Genome Project more than a decade. What arrived in the mid-2000s changed one thing: instead of reading one fragment at a time, it read millions at once, same chip, same run.
Do in parallel what you used to do in series. That is the whole ballgame, and it is the same jump that turned one bank teller into a wall of cash machines, or one scribe into a print run. Nothing about the single read became magical. There were suddenly millions of them at the same time.
I learned all of this the way a salesperson learns anything, slowly and in public. Most of what I know about genomics I picked up selling those platforms, and I should say that plainly, because I do not sell them any more.
What the job actually gave me was rooms. An Illumina conference in Bangkok is where I first met SOPHiA GENETICS, who we represent today. On 8 December 2009, CSIR-IGIB announced the first human genome sequenced in India, read in fragments of 76 letters, millions of them at a time. The announcement came ahead of the peer-reviewed paper, which drew some criticism at the time and arrived in 2012. Some years after that I was in a room with two of the scientists behind it, following perhaps half of what was said and writing the rest down to look up later.
The rooms are the reason I can follow the conversation at all.
Here is what parallelism actually bought, as far as I have watched it.
A baby in a neonatal intensive care unit. A lab we support returned a whole genome, sequenced and reported, in under 48 hours. The result changed what the team did next, and what they could tell the parents.
Parallelism made that possible and did not make it easy. Almost none of those 48 hours went on sequencing.
Reading the genome is the fast part now. The hours went on library preparation, on compute, on somebody curating variants against a database, and on a clinician being reachable at 03h00 to act on the call.
The Solexa acquisition dates and value are from Illumina's own announcements of November 2006 and January 2007. The first Indian human genome announcement, the date and the 76-base read length are from CSIR-IGIB; the peer-reviewed paper is Patowary et al., Human Mutation 33:7, 2012. The 48-hour turnaround is ours, observed in a lab we support, and is not from any published series. Published work on rapid whole genome sequencing in intensive care, including Project Baby Bear, reports time to diagnosis rather than turnaround, and reports it in days.
On what the fast part is not: The sequencer is the easy part.