When Software Eats Bio | Hacker News Reader

When Software Eats Bio

a16z.com

43 pointssrunni15 comments

One area that's ripe for disruption is Next Generation Sequencing (a high throughput derivative of Sanger sequencing)

Illumina produces devices called sequencers, which are glorified microfluidics + imaging devices, which uses base pair chemistry to build the dna complementary strands of single stranded dna fragments one base by one using special dna bases that have washable fluorescent die and are blocking (so that the complementary strand grows by one base). Once the correct dna base is stuck to the single strands on the glass device, a picture of the glass device is taken by shining laser and using a CCD imager. The Nature of DNA makes sure that the correct complementary base gets stuck. 4 TIFF files are generated with billions of tiny dots one for each base A,C,G and T. The fluorescent die is washed, and the last added ddna base is deblocked and the cycle repeats. The whole software suite (image processing of TIFF files) and hardware specs (fluid chemistry, etc) is unfortunately closed source :(

I think it's more accurate to say that NGS is still in the midst of disrupting biology in general. Illumina has been consistently dropping sequencing prices for the last 7 years. Oxford Nanopore is a very promising technology as well that could completely disrupt the space.

I wouldn't say the the industry is ripe to be disrupted by software though. The cheapness of sequencing enables the development of new software for analysis, which in turn enables more sequencing.

The next frontier in NGS is long read SMRT (Single Molecule Real Time - PacBio) combined with Oxford Nanopore's technology. That is, very long reads combined with massively parallel transcription that doesn't rely on light emission. Illumina-based tech is not going to be part of the equation.

Illumina is really only related to traditional Sanger sequencing in that its sequencing by synthesis.

There is no way that this is going to be the case. It's much more likely to be a combination of Illumina and ONT or ONT alone. PacBio is far too expensive and low throughput to be a challenger outside of niche applications (like genome assembly).

I would much rather have ONT backed by Pacbio than ONT backed by Illumina. ONT is well suited for doing WES quickly, with better fidelity than Illumina and PacBio is much better for supplemental data to do genome assembly using reads from a short read system

What evidence do you have for ONT having better fidelity than Illumina in any application? Illumina's error rates are typically <= 1% while on "good" reads ONT has error rates cited at 15% and overall is even worse[1]. I agree that PacBio is better for genome assembly, but for many cases we do not need to do assembly. Resequencing workflows with alignment and variant calling are pretty good.

1. http://www.sciencedirect.com/science/article/pii/S2214753515...

I'm extremely skeptical that ONT's sequencing will become cost effective. Library preps for long range correlation + Illumina will likely dominate; why not get high fidelity + long range off the same instrument and reagents? Dovetail and 10X are two companies doing this right now.

That or a different nanopore tech. After all the time they've spent and the missed promises of ONT's CEO it doesn't look like the current generation will deliver.

Disappointed no one brought up the regulatory hurdles involved when discussing the resistance of the industry to disruption. When the cost to bring a new drug to market is over 2 billion dollars [1], it's hard to see start-ups competing in that environment.

1: http://www.scientificamerican.com/article/cost-to-develop-ne...

Considering that a16z said that they are focusing on software around biotech and not on traditional biotech I think it makes sense that they aren't talking about regulatory hurdles or the cost to bring a drug to market. Eroom's law is a serious problem, and it is hard to imagine a disruptive startup that could compete in that environment.

It sure is an interesting time for melding computational advancements with medical use cases. The reason that the cost decay can promulgate the previously missing disruption seems to be missing the bigger question of regulation in these fields.

The real question is how they will filter for Theanos like companies

Theranos was transparent about the overhyping from the very beginning to somebody with an understanding of bio. It's more shocking that so much of the "tech" media bought it.

Imagine a bunch of old-school bankers or Fortune 500 CEOs from the 90s trying to pick a winning software business. They know "business" so they should be able to apply their knowledge to the software world easily, would be their reasoning. Silicon Valley's picks in biotech, like Theranos, have fairly transparently misapplied "knowledge" in the same way. They look for all the wrong things, attributes that work well in software but not other places: 1) confidence bordering on arrogance, 2) dismissal of experts and standard knowledge disguised as an attempt to "innovate.", and 3) a chance to "hop on" to a rising founder/CEO before they've proven themselves and become super expensive. Similarly, be skeptical of other SV health efforts (such as Google's) unless there are real scientists behind it and also real experience in biotech (not tech).

Complete Genomics (next to LinkedIn and Google) is an example of what happens to a biotech company that has really great tech, but runs the business like a tech company rather than a biotech company. Scale goes completely sideways, they miss the market, and get sold off to a bigger fish and most likely will languish.

Can somebody talk about CRISPR and particularly business ideas around it? I read the recent New Yorker article (http://www.newyorker.com/magazine/2015/11/16/the-gene-hacker...), but still have a lot of questions!

http://www.technologyreview.com/news/543181/crispr-gene-edit...

Check out http://desktopgenetics.com - web based crispr design app

Reply on news.ycombinator.com