SoftBank is notorious for some really bad tech bets, particularly WeWork, a company that suffered from dual bright red flags. WeWork was providing co-working space, which is definitely in demand as many talented people strike out to be entrepreneurs and need professional space but require many expensive amenities (such as conference space) only occasionally. But WeWork's insane valuation - -it was going to IPO IIRC with a market cap approaching $20 billion -- was despite being in a field with virtually zero barriers to entry. Based on news reports, WeWork's primary strategy for growth was to use their capital to grossly underprice their space and drive out competitors, presumably planning to ultimate jack up rents once they had monopolized the market. But what would stop another player from running the same strategy?
Worse, the CEO and his spouse not only treated the company like their personal piggybank for all sorts of peripheral projects, but was also cutting friendly deals with himself to siphon assets that might have value away from the company and into their own personal ownership. SoftBank was going to proceed with the IPO and further shoveling of money into this furnace even after it was widely reported in the news. Not exactly the most discerning of investors.
For PacBio, with strong management (which, to remind you endlessly, includes their CEO who sits on my employer's board) will presumably never suffer from SoftBank attempting to micromanage their growth plan -- let's hope there is no 180 degree turn in attitude on how carefully they should watch how their money is spent!
How might PacBio productively apply this cash?
There's so much to be done on this platform, not because PacBio isn't a good platform but because everyone in this space must run a Red Queen's race. With more funding there is opportunity to really push the platform. I've covered some of this before, but let's review some of the potential areas for high activity -- which of course takes the form to a degree of "how would KR spend PacBio's money if he were there?". Well, how besides showering some on loyal bloggers.
Even with so many financial resources, it is still important to spend them well. A key to that is focus: PacBio IMHO should remain focused on providing high accuracy long read sequencing technology to be applied to a wide range of life science challenges and anything not directly supporting that should be given a pass. As a historical reference, Illumina has sometimes flirted with tools not on the sequencing path -- remember the Eco line of qPCR instruments? Sure, qPCR can be used to quantitate libraries, but isn't core to being the best short read sequencing technology company.
A corollary is that it is important to create a rich portfolio that diversifies activities across many dimensions. For example, it is important to have some balance of projects that might deliver small gains quickly versus those with huge impacts many years down the road.
Clearly core is improving the performance of the instrument on both the hardware and chemistry. This new cash cushion should enable a concerted push for higher density SMRTcells and the optical hardware to support them. That would enable more reads per flowcell, which would drive down the cost of big projects and close the cost margin between short read and long read clinical genomes.
Going for faster polymerases and higher frame rates would be another big swing, possibly big miss approach which a well-funded company can risk. Since photodamage to polymerases is still a major termination mode for the HiFi reads, getting more data sooner could mean more passes across inserts and enable both longer inserts in HiFi mode as well as higher accuracy single molecule consensus sequences, not to mention better instrument utilization.
Even tweaking the current chemistry by small amounts could make customers significantly happier. Loading efficiencies are apparently typically in the 40-50% range. On the one hand there is at most about 2X room for improvement here, on the other everyone loves getting more yield from the same instrument run. Any small improvement in reducing photodamage can be significant when multiplied by millions of sequences.
There's also the little things. For example, how can you QC a library before devoting a sequencing run and precious instrument time to it? I saw an interesting assay in development for this a few years ago; don't know what happened to it. Such an assay might also be the first part of a screening funnel for changes in chemistry, enabling medium to high throughput screening of different polymerases or sample conditions.
Library prep is an area where PacBio has innovated extensively, but there's always room for more improvement. The Ultra-Low kit uses amplification to enable sequencing from minute amounts of input -- single digit nanograms. But the trade is that the library diversity is reduced so it isn't suitable for large genomes. Further optimizing the protocols to drive the input requirements for the standard (with no genome size restriction) and Low Input (medium genomes acceptable) could be very valuable. If it is obtained by reducing the amount of library that actually must go on the sequencer, all the better.
PacBio library preps are also relatively labor intensive processes, with ligation and bead cleanup steps. Getting these routinely automated will be important to push into high throughput applications.
PacBio was early on calling base modifications, but has sensitivity issues with 5-methyl-C
modification. A Taiwanese group recently announced that they have a machine learning approach to fix this; PacBio should probably license it asap. They should also look at moving methylation analysis into HiFi mode and onto the on-instrument compute that comes with Sequel IIe. Not only would this be more convenient for customers, but it would enhance the ability to detect haplotype-specific methylation since high confidence methylation calls could be made on single molecules.
DNA extraction aka sample prep remains an area in need in of further innovation and automation; too few extant protocols are automation friendly. Since HiFi is concentrating on long but not ultra-long DNA (25 kilobases), the challenge is eased a bit, but there is still room for improvement. This is an area worth many parallel efforts, as the prize is big and the path to it non-obvious.
Related to that, there are startups in the DNA extraction space and there's probably more interesting technology leads buried in academic labs. PacBio should think carefully about a strategy involving venture-like funding, partnerships and outright acquisitions across their platform, but it probably applies particularly in the DNA extraction space. I differentiate funding from partnership in the degree of involvement -- for funding PacBio would primarily define requirements and goals and hand out some seed money, whereas partnerships (which the earlier stage would evolve into if successful) would mean a close collaboration between the outside company and PacBio scientists.
Acquisitions are tricky and an area where I have decidedly mixed feelings, having seen several up close. It is very challenging to merge the cultures and workstreams of two previously independent organizations, and mergers often lead to much unproductive chaos, disruption and uncertainty. But if correctly executed they can weld complementary teams together or meld them into sums greater than the parts, and can also lock in exclusive access to key technologies. I'd mostly advise going slow -- trying to slam two groups together is the riskiest -- though I've seen it pulled off well.
A related way to spend cash a bit freely but wisely would be to set up an applications lab devoted to relatively short term customer collaborations to extend the platform. Have a difficult sample type? Bring it to the applications lab! Have some crazy idea to apply the technology? Ditto!
Similar to this, PacBio should think about broadening access to the technology and particularly in training the biotech workforce on the platform. Very few educational institutions will have a Sequel IIe and even fewer will let undergraduates near one. If there was a "PacBio U" which enabled undergraduate or Master's students to learn how to use the instrument, the pool of trained PacBio operators would greatly expand. Such a program should be structured and marketed to ensure that students at a wide slice of schools -- not just glam schools and state flagships -- have access to such training.
Such a facility might also spend some weeks training operators for companies that are planning or contemplating an instrument purchase. That should be part of an overall effort to continually review what the pitfalls of instrument installs are -- these are complex instruments with many requirements and it is hard to believe they are as plug-and-play as a new coffee machine.
Of course a huge area to invest in is software. A small bit here would be to update all the workflows to work from HiFi data -- there's not much reason to not go all-in on HiFi. But a far larger task will be to build tools so that large centers can efficiently operate vast fleets of Sequel IIe instruments.
PacBio operates in a number of interesting market segments. I would define the most valuable ones as human genomes for clinical use, plant genomes for agricultural use, human biotherapeutic QC (such as adenovirus vectors), large genomes for agricultural and aquacultural breeding, synthetic biology, metagenomics, infectious disease, and industrial microorganisms. Plus a general category of supporting basic research and whatever doesn't fit into any of those. Most of the improvements I've outlined will apply well in teach area, but there will certainly be some differences. Clearly human clinical genome software will have many regulatory p's and q's to adhere to faithfully and human biotherapeutic QC would also intersect with a regulatory sphere. That one could be particularly valuable if PacBio could convince the FDA that it is uniquely suited for verifying the integrity of production lots for both length and sequence. Some of these segments may have specialized DNA extraction challenges.
In any case, it should be lots of fun to watch -- PacBio hasn't been afraid to blaze trails in the past nor were they paranoid about enabling competitors -- quite a bit of valuable tool development spurred by PacBio ended up being used on Oxford Nanopore data. In the end, I think that benefited PacBio more than it hurt them -- too many companies have stifled their own user base from improving a platform by excessive secrecy and information hoarding.
Wanted to get this out before I completely switch my spatial attention to AGBT -- the next several days should keep my fingers busy!
"nor were they paranoid about enabling competitors"
ReplyDeleteApart from sneaking cellphone pics of academics' nanopore slides and posters, that were presented at a closed door conference, to then file patents on and sue the competition with you mean?