The Futility of Defining Sequencing "Generations"

[AllSeq]

I just recently answered a question on Quora about the number of “sequencing generations” there are. (My answer was basically that the concept of generations for sequencing has never been used consistently, doesn’t make much sense and isn’t very helpful.) In doing so I quickly categorized the various technologies based on a number of attributes:

1) sequencing-by-synthesis vs direct sequencing (vs ligation vs hybridization)
2) stepped vs continuous sequencing
3) optical vs non-optical detection
4) short vs long reads
5) single molecule vs amplified

And here’s how I mapped the major platforms to these categories:

Illumina = SBS, stepped, optical, short read, amplified
Ion = SBS, continuous, non-optical, short read, amplified
PacBio = SBS, continuous, optical, long read, single molecule
Oxford Nanopore = direct, continuous, non-optical, long read, single molecule
Helicos/SeqLL = SBS, stepped, optical, short read, single molecule
454 = SBS, continuous, optical, short*, amplified (* considered ‘long reads’ at the time, but now at least an order of magnitude shorter than the long read technologies)
SOLiD = ligation, stepped, optical, short read, amplified

I’d be curious to hear what others think. The least satisfying is the “stepped vs continuous” category. ONT doesn’t fit perfectly cleanly as they could be considered “stepped” since a ratcheting protein is used, but it doesn’t have the distinct cycles of Illumina’s SBS chemistry. Also, I wasn’t 100% sure about the Helicos technology – is it cycle driven or continuous like Ion? Thoughts?

[SNPsaurus]

Helicos is one base then image, like Illumina.

I think nanopores are continuous. I don't see much of a difference between a polymerase adding nucs and a template being moved through a pore by enzymatic action. Certainly the timescales involved group Oxford Nanopore with other continuous platforms.

Nice way to organize them, though!

[AllSeq]

I don't see much of a difference between a polymerase adding nucs and a template being moved through a pore by enzymatic action.

That's a really good point!

[gringer]

You missed out sanger sequencing (e.g. ABI 3730), which doesn't really fit into the nice bins you've created. My best guess:

SBS, stepped, optical, long, amplified

Are there any electrical (or non-optical) sanger sequencers around? Would radiosequencing count?

[GenoMax]

You should consider a "time to sequence" component.

We went from day(s) (if you include sub-cloning needed etc) in case of Sanger (first radio-labled then ABI) to less than a week (454, initial Illumina) to back to days (2000,2500), <24h (PacBio per SMRTcell, Ion per chip, not counting prep) and back down to a couple of hours (nanopore).

[gringer]

For pure detection only, nanopore can give you a diagnostic indication within a couple of minutes (assuming a great flow cell, great sample prep, and high proportion of target genome in the sample) -- It doesn't really make sense declaring a particular length of time for sequencing. Maybe 'interruptible' would be better.

[AllSeq]

You missed out sanger sequencing (e.g. ABI 3730), which doesn't really fit into the nice bins you've created. My best guess:

SBS, stepped, optical, long, amplified

Are there any electrical (or non-optical) sanger sequencers around? Would radiosequencing count?

I specifically avoided Sanger sequencing, partly out of laziness and partly because people usually refer to 'generations' in 'next generation' sequencing. That said, I would agree with your assessment apart from 'long' - PacBio and ONT have really redefined what 'long reads' mean! Also, I think radiosequencing would still count as 'optical' as film is used as the detection method.

[AllSeq]

You should consider a "time to sequence" component.

We went from day(s) (if you include sub-cloning needed etc) in case of Sanger (first radio-labled then ABI) to less than a week (454, initial Illumina) to back to days (2000,2500), <24h (PacBio per SMRTcell, Ion per chip, not counting prep) and back down to a couple of hours (nanopore).

Adding "time to sequence" is an interesting idea, but it seems complicated. I was keeping the list at a pretty high level - e.g., Illumina vs PacBio. The time to sequence would vary depending on a specific platform (e.g, MiSeq vs HiSeq X) and would vary from application to application. But it is definitely an important variable.