Over the last few years, I have seen sequencing companies show that the cost per base is being reduced at a rate faster than Moore's Law. Well, now that means that this law does not apply!
Is there a new law or rule of thumb that can better describe the cost per base over time? Any mathematicians want to give this a go?
Moore was able to generate his law because he was seeing a linear path which the processors were following: it was a matter of how many transistors you were able to put in the same silicon base. the idea was that this evolution was fairly easy to study and to extrapolate, although as Niek said there are ways to increase efficiency in orders of magnitude without having to increase necessarily in the same order the amount of transistors per processor (parallelism mainly). the reason why the typical graphs of sequencing costs were following more or less Moore's law was precisely because the costs were being reduced because the existing technique (Sanger sequencing) was evolving linearly, until parallel sequencing appeared. we would have to see a comparison of processor prices with multicore technology considered in order to compare sequencing and computing power appropriately.
the point I wanted to raise is that the sequencing costs drastically reduced when parallel sequencing appeared, and that's a major change in the technology. and the problem to estimate any evolution of these cost without realizing that sequencing technology may change drastically too in a couple of years time (3rd generation sequencers, or who knows what) is not very wise. anyway, if you don't consider any evolution in the sequencing technology to happen in a few years time (I very much doubt it), I would point out that the current sequencing costs do still follow Moore's Law now (during last year at least) in terms of slope, after evolving almost reversely exponentially for a few years. the only difference with Moore's law's slope is "simply" a gap which the technology evolution represented, and I would say that this slope will be maintained until a major change on the technology happens. something that, I should say, should not surprise us to happen in 2 or 3 years time.
Moore's law is about the amount of transistors doubling per 2 years. There are other ways to lower cost than doubling computer power, which is why the law doesn't apply. One thing that severly lowered the cost was the new way of cloning, using emulsion pcr instead of the old pcr. Also, the reads that are being sequenced have increased tremendously, lowering the computing cost for assembly. And now with 3rd generation sequencers they can read the bases in real time.
Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
version 2.3.6
