Wednesday, January 22, 2020

Genia: It's Alive!

Once in a while I get the thrill of someone sending me a really good tip - that surge that comes with knowing that I know something that most people don't.  That rush of knowing that soon I'll get to spill the beans.  It's great, even if it upends my widely disseminated opinions.  Even if a moment later I realize that if I had thought harder I would have unearthed the nugget on my own.  All of which is the case here -- my tea leaf reading that Roche partnering with Illumina on diagnostics indicated that Genia is no more -- well, that interpretation is no more.  Because I got a hot and verifiable tip that Genia is very much still an active project at Roche.  And the verification is how I could I found this independently.

The tip was simple but electrifying: my correspondent knew someone who had been recruited recently to work on Genia.  Indeed, they had been recruited as well and had some further information: Genia is intended to be integrated into a complete sample to results system.  Verifying the hiring bit turned out to be easy -- just go to Roche's jobsite and search for "nanopore" ("genia" did not work as a search term).  That currently pulls up 7 positions, and nicely the search results page says when each was posted so we can see "Principal Data Scientist, Bioinformatics" posted this week and "Scientific Manager, Nanopore Engineering" posted at the beginning of the month.  The other postins are from November, September, August, June and May.  All but one are for the Santa Clara CA site; one bioinformatics opening is for South San Francisco -- that turns out to be in the Genentech group and so a red herring.  

But the others have job descriptions leading with sentences such as this for the Scientific Manager position:
Roche Sequencing Santa Clara is developing nanopore-sequencing technology to improve the lives of people in a wide variety of applications, including the timely screening of newborns for diseases and the diagnosis of cancers and infectious diseases. Roche is looking for a motivated and team-oriented individual who is passionate about finding new ways to address these urgent needs. As a Scientific Manager, you will be leading scientists that are designing, expressing, purifying, and characterizing biological nanopores for Roche Sequencing Solution’s (RSS) next generation sequencing platform. The successful candidate must be excited about single molecule kinetics and learning new concepts from chemistry, physics, algorithm development and electrical engineering and applying them to an interdisciplinary project.
For a Bioinformatics position:
We are seeking a self-motivated Principal Data Scientist, Bioinformatics to join our Nanopore Sequencer Data Science group, where we collaborate with experimentalists and algorithm developers to build the Roche single-molecule DNA sequencer. 
Or for a GPU Scientist position:
Roche’s proprietary nanopore-based platform allows for single molecule, electrical, real-time analysis without the need for complicated optics or fluidics. Developing our unique platform requires expertise across a range of fields: software engineering, data science, biology, chemistry, protein engineering, circuit design, and electrochemistry, just to name a few.  We develop a proprietary system that generates data at tremendous rates, a big challenge for existing technology for data processing.
One of the bioinformatics positions isn't strictly dedicated to Roche's nanopore platform
The work will be focused on developing novel applications for sequence data from multiple technologies, including the Roche nanopore sequencer.
Or for a Deep Learning Scientist
The Deep Learning team is responsible for architecting and training models using terabytes of proprietary DNA sequencing data, towards the central goal of providing accurate “base calls” -- i.e. translating our raw voltage signal into DNA letters. 

So Genia is very much alive.   They're treading some obvious paths given the technology: improving the pores, exploring different membranes and using deep learning to make basecalls -- as well as getting everything to behave when slammed together into one device.

I've actually been poking at some startups' job sites to see what information can be gleamed from their job postings.  For example, Omniome is looking for Marketing Manager and Reagent Manufacturing and other positions that go with commercial launch.  They haven't announced such a launch yet, but that and a recent $60M fundraising suggest that it isn't far away.  Other companies that might be getting close don't have similar job postings.  Anyway, that's a topic for another post.  But job postings are one way companies almost without option leak information to the outside world.  Now, it isn't a guaranteed source -- a company with a stable workforce won't have any ads and some small companies may work almost entirely through networking (this was mostly the case early at Warp).

Why is Genia so overdue?  One possibility is that despite two papers showing general proof-of-concept (which I covered as they published in April and October of 2016), actually making a working device has proven elusive.  But another is that the perfect became the enemy of the good; instead of getting a workable sequencer-only module going early they've held out for the whole enchilada.  Roche has said they don't plan to enter the general sequencing market but instead focus on a vertically integrated platform for clinical use -- the siren song that lured QIAGEN onto the rocks and now is entrancing Agilent. Can you tell I'm not a fan?  It's easy to see how one might convince themselves that the way to win in clinical is a fully integrated system -- but you can't win with a system that never hits the market.

Got a good tip?  Want to encourage me to dig in a particular direction? Shoot a note aimed towards keith.e.robison at (but use the single letter code!).  Or message me on Twitter.  Or leave a cryptic comment on this piece!


Anonymous said...

Thanks for your Blog.
Last news (with reference to Genia):

Anonymous said...

I don't think Genia is alive. They also lost their IP battle with UCSC. a rump effort remains at Roche on Nanopore sequencing, my bet is not closely based on the Genia stuff.