Important Events in the DNA Industry and Bioinformatics, December-January 2018-19
Due to the winter holiday lull there was less business news from our industry than usual. However, there is one news item that is remarkable because it allows for comparison of the progress of development of DNA sequencer startups with analogous next-generation technologies. In early January, Roswell Biotech announced that it had attracted $32MM in Series A financing. Roswell has developed a platform that enables real-time tracking of polymerase processes integrated into nano-electronic circuitry. According to this source, Roswell's previous rounds of investments for the period 2014-2018 total about $6MM. During this period, based on the statement of Roswell's president and CEO about the company's future plans, they (among other things) created sensors and developed a molecular electronics chip.
Gamma-DNA recently published a report on the work it has completed over the past year. One of the most significant events as such is not only the creation of a prototype sensor, but also the time frame in which it was created as well as the related costs. The development period was less than half a year and the cost was about $40K. In less than six months, we will demonstrate 170-nucleotide-long sequencing of an oligonucleotide on a discrete sensor. Importantly, Gamma DNA is creating not only an efficient technology, but also an effective and convenient tool for B2B users with the instrument's capability to progress depending on the market needs. We have no doubt that other companies are also working in this direction and always welcome reasonable openness about achievements, as for example, was done at Roswell. We also hope that our last year's report will be useful to other DNA sequencer developers.
Therefore, starting this year in our news reviews we will make comparisons between our technological development and that of competitors whenever there are obvious and objectively comparable situations. As in the IT industry with their AI and deep learning benchmarks, as in here (Google vs Nvidia vs Intel hardware benchmarked for deep learning training), it would be great to see – and we would readily welcome that – similar performance benchmarks in the field of DNA sequencing too. Such metrics will allow users of DNA sequencers to determine more easily and quickly which sequencer is most suitable for them. The manufacturers themselves would be able to gain insights into what part of their product needs improvement or what areas need smarter resource allocation. The world has seen not only drastic technological changes, but also the benefits of the emergence of joint business solutions as a result of interaction between competitors. Note: the above link is a platform of AI benchmarks jointly formulated, coordinated between and submitted by the top players in the market of software and chips for AI and machine learning.
Stay tuned for more news!
Important Events on the DNA Diagnostics Market, November, 2018
It’s been known for a while now that short-read sequencing technology maker Illumina has acquired Pac Bio, specializing in the long reads, for $1.2B. If approved by regulators, the deal is expected to close in mid-2019. It is vital for the market to appreciate the incentive behind the deal. According to Illumina’s president and CEO Francis deSouza, it was access to Pac Bio’s new product line. With the help of Pack Bio’s new 2019 chip, it will be possible to sequence long DNA fragments faster and cheaper. Combining the technologies of both companies will allow creating more attractive products with superior precision and speed.
This deal is happening in the times of the so-called technological shifts in DNA sequencing. What does this shift mean? The market is moving toward long reads. The long reads do not require reference data with which to compare the resulting genome to achieve benchmark accuracy. At the same time, long reads allow for the extraction of more new knowledge and offer more profound insight than short reads provide. As an example, Illumina’s short read is 150 base pairs in length, while Pac Bio’s is 20,000–30,000 bp (200,000 bp is the current maximum). In one of his interviews two years ago, Evan Eichler, a prominent genetics professor at the University of Washington and one of the Human Genome Project's participants (1993–2000), stated that as soon as the cost of full genome sequencing from long-reads drops to no more than double that of short reads, the days of short-read sequencing will be numbered.
According to Pac Bio’s October 2018 investor presentation (slide 12), this time will come at the beginning of 2019, when their long-read full genome sequencing is expected to drop to $1,000—the cost of Illumina’s short read sequencing. That being the case, Illumina is concentrating on switching from long-read to short-read sequencing, which will ultimately bring the cost of full genome sequencing down to $100.
The long-read market is also powered by Roche and Oxford Nanopore (the length of the latter is 2 million bp). The market believes Oxford can surpass Illumina in sequencing precision. According to Clive Brown, Oxford CTO, thanks to an AI/machine learning application, Oxford has overcome the hurdle of low accuracy in nanopore sequencing. This is the reason for the technological shift toward long reads and the motive behind the Illumina-PacBio deal, as hypothesized in The Motley Fool article. Incidentally, The Motley Fool is one of the few finance companies that managed to advance to create its own venture fund, MotleyFool Ventures, and raise $245M from its current customers.
As the cost of full genome sequencing of $100, on par with competitive accuracy of cheaper devices, is poised to become an industry standard, it is useful to observe how every manufacturer of DNA sequencers will be able to differentiate itself from others in the long haul. The more so as buying a small gadget or industrial equipment is not always influenced by the price or a feature. But first, we need to answer the question on how consumers (clearly not B2C) purchase sequencers and what considerations determine their choice. The answer: The selection of a sequencer depends on the type of task and accrued experience of solving that task using various sequencers. First-time buyers purchase a widely known brand which might not necessarily fit the buyer in addressing a particular task.
Thus, the more customizable a sequencer is, the more attractive the brand will be for a consumer. It’s important to understand that the existing generation of users has reached a critical mass to influence the choice of first-time buyers. When replacing a sequencer, some users will prefer their current brand while others will choose based on their new knowledge and previous experience.
Should the makers of sequencers want to anticipate user preferences, they will need to receive reliable feedback from customers. The quality of the feedback depends on trust, which, in turn, requires time to build. The sooner the company establishes trust with existing users, the more time it will have to find ways to attract new customers with more tailored equipment.
When a new or improved product is released, a community of experienced, trusting users will be able to evaluate the new sequencer and issue recommendations. For new consumers, such recommendations will serve as grounds for believing in a new product without a proven track record. For the maker, the feedback will provide useful data on what needs to be improved. For established companies, it will be a way to grow, and for newcomers, it will offer a good start.
To achieve this model, the makers will need the following things: (i) user experience data broken down by task/application and (ii) a technologically flexible platform (not settings in the usual sense) that will allow users to configure the sequencer according to their experience, knowledge, and current tasks. User experience data (i) is the starting point for any maker to set its product apart from competitors in a sea of standardized equipment or as a new maker entering the market and lacking a reputation.
Stay tuned for more news!
Important Events on the DNA Diagnostics Market, May-mid July, 2018
Illumina has purchased Edico Genome for $100 million. Edico Genome is a developer of Dragen - a Field Programmable Gate Array (FPGA) based platform that implements genomic algorithms. It will be curious to watch if Illumina uses Edico in the development of their own dedicated neural network processing unit (NNU). A dedicated NNU will allow data processing in their sequencing machines utilizing edge computing thus eliminating the need to send the work off-site. It certainly would be a go-to tool for running quick tests that retail customers will find convenient.
Illumina is also actively expanding partnerships with oncology research companies. Its new additions are Loxo Oncology and Bristol-Myers-Squibb. Illumina's own research arm, Grail, is focusing on developing ways to detect cancer at the earliest, presently undetectable stage. The basis of the technology, naturally, is genome sequencing from a patient’s blood. Samples taken from 580 patients without diagnosed cancer returned five positive results, and of those five, two had cancer diagnosis confirmed.
Consumer genomics continues to gain traction. Genealogy tests and sensitivity to caffeine tests experienced 44% growth in services. These tests lend themselves well to the business of beauty salons, spas, and fitness centers provided the companies have stationary or portable DNA sequencers in-house.
LifeBit raised $3 million in seed round capital. LifeBit is developing a cloud and machine learning-based software to allow customers to gain actionable insights into their sequenced data. Incidentally, LifeBit does what an NNU can do with a partial or full set of sequenced data using edge computing. LifeBit represents a market segment of sequencing-as-a-service, similar to a model in which businesses outsource their IT tasks to other tech companies.
More news on the way, stay tuned!
Information letter: some important events on the DNA diagnostics market, September 2018
Information letter: some important events on the DNA diagnostics market, August 2018
Two recent developments in the interface between technology and biological sciences create potential solutions to the problems of incentivizing donors to provide DNA samples for medical research and separately to the development of advanced neural networks. The first problem is resolved by using blockchain technology to encourage customers to share DNA samples. The second problem uses the behavior of certain genes and the knowledge about such behavior to create spiking neural networks (new generation networks that differ from the existing generation in closely emulating neuron network behavior of living organisms).
Companies such as 23andMe and Ancestry augmented their user agreements to disallow the release of customers’ DNA data to Big Pharma companies without consent. According to 23andMe, 80% of their five million customers agreed to allow their data to be used for research by third parties. One of the 23andMe customers ironically asked: “If my data has so much value that Glaxo (GlaxoSmithKline is 23andMe’s partner) can use it to create a new drug, then what can you offer me besides the finding that my eyes are most likely brown?” Customers who agree to release their data to third parties receive no benefit other than the satisfaction that they are contributing to medical research. The use of incentives tied to reimbursement via secure technology such as blockchain can materially increase the customer base of participating companies and, correspondingly, accelerate pharmaceutical innovation.
An example of such collaboration is the use of photo and video content created by customers of large technology firms which is then used by those companies in training their ML models. It became possible due to progress in machine/computer vision. Likewise, significant progress in new drug development may arise from correct incentive programs that allow DNA sharing through safe instruments. Sequencers might become such an instrument as part of a blockchain being in effect co-computing devices and nodes authenticating tissue sample ownership. A customer would pay for his DNA test and get reimbursed for sharing the results.
More news: нas Gattaca time finally arrived? New research revealed 1000 gene variations that influence how far a person progresses through school. The research is based on the study of a gene that encodes a protein in a person’s neural network.
More interesting than the direct result of this research is how this research, alone or with other findings, may be continued.
As is known, data in artificial impulse neural networks, as in biological ones, are formed and transmitted by a series of impulses, the main characteristic of which is their dynamics.
For example, two series with the same number of impulses but with different time intervals between them represent two different types of information. Each series of impulses is an individual piece of information or an element of behavior. The study on the correlation between genetic makeup and school attainment aims to determine how certain genes influence the formation of behavior patterns. In other words, is there an interplay between specific genes and how neurons form a series of impulses. Discoveries in this research will undoubtedly aid the development of impulse neural networks useful in commercial applications.
The majority of current research on impulse neural networks does not take into account the DNA component. A new computational model of a neuron is breaking the mold. In the existing model, the single threshold unit summarizes incoming signals and gives a “green light” to a spike if a threshold level is exceeded. The new experiments demonstrate that a neuron contains multiple independent threshold units with unique entry signals and threshold values for generating a spike. Aside from overturning the established “neuro-norms,” the DNA component might be the missing link in the quest to bring energy efficient spiking neural networks to market.
Stay tuned for new developments!
Important News on the DNA Diagnostics Market Mid-July - Early August 2018
Quantapore, a developer of nanopore-based nucleic acid sequencing technology, raised $15.6M. Five funds participated in the round: Northern Light Venture Capital, Tsingyuan Ventures, Sangel Venture Capital, Baidu Ventures, Cloudstone VC. The developing DNA sequencing sector is constantly adding new tasks that cannot yet be implemented in a single technology (sequencer). The current practice of parallelization (using sequencing machines with various underlying technologies from different manufacturers and switching between them depending on the task) allows for some technological and economic flexibility. However, this flexibility is achieved through an “extensive” mode, meaning that to solve even their local tasks sequencer users are forced to buy several models from different manufacturers. In this case, a user needs to be able to handle parallelization and be versed in the related nuances of combining different technologies and manufacturers.
Additionally, the necessity of buying multiple machines increases costs. This is unduly burdensome to hospitals and other B2B users. Mass customer appeal requires a convenient user experience including the ability to plug in the device, press the button, and tweak the accuracy/speed/cost settings without the need for interfacing with different technology or machines. When evaluating DNA sequencing startups and new announcements made by the entrenched incumbents, it is essential to identify if their products can become a true one-size-fits-all solution.
The participation of Baidu Ventures in Quantapore is noteworthy even if it is passive. BV is a unit of the Chinese Baidu from the so-called Super 7 Group (Google, Facebook, Amazon, Microsoft, Alibaba, Tencent, Baidu). Baidu, similar to the other giants from this group, is one of the leaders in AI computing with their Apollo autonomous driving platform, and in their developments in DL compute acceleration on GPUs, FPGAs and custom hardware. Baidu only invests in AI/ML/DL startups. In this particular deal, the participation of BV speaks volumes: it is either about the insights Quantapore has into the development of compelling soft or hardware-based processing of sequenced data with neural networks, or it is about the opportunity with one of its portfolio companies which builds an FPGA accelerated solution for genomics (similar to Illumina+Edico Genome). If BV’s decision to participate was based on other opportunities which might be just as promising as DNA sequencing, then one can expect to see other tech giants follow suit who are not currently involved in DNA sequencing.
Another company, Omniome, raised $60M in funding. It’s unclear whether or not they are going to manufacture their own sequencing machines. Omniome emphasizes the polymerase’s natural matching ability when it comes to obtaining unprecedented precision.
Exciting studies conducted by a Princeton University lab demonstrate the time series behavior of an enhancer and its interaction with its target gene. Enhancers, the so called ‘junk DNA,’ were once thought to be useless DNA that doesn’t code for protein. An enhancer is a sequence that ultimately controls the copying of “original” information from DNA to messenger RNA. According to the video made by the lab, the enhancers are responsible for not only switching on the function of their target genes but also keeping it active as long as they are attached to their genes. Once an enhancer got detached, the function gets switched off. Gaining more insights into this interaction and the better understanding of an enhancer’s behavior is important in finding the ways to ensure successful “re-engineering” of an organism when preventing or curing diseases.
Stay tuned for more upcoming new developments!
Information letter: some important events on the DNA diagnostics market, May - mid July 2018
Illumina purchased Edico Genome for $100M. Edico Genome is a developer of Dragen - a Field Programmable Gate Array (FPGA) based platform that implements genomic algorithms. Will be interesting to watch if Illumina is going to bring Edico to bear for the development of their own dedicated neural network processing unit (NNU) so their sequencing machines become able to process the sequenced data at the edge without the need to transmit it to the outer world for further processing. It certainly would be a go-to tool for running quick tests that retail customers will find convenient.
Illumina announced collaborations with Loxo Oncology and Bristol-Myers-Squibb to develop companion diagnostics in Oncology. Illumina’s Grail is on the upswing. Grail is developing the ways, based on data sequencing, to detect cancer at the earliest stage that is currently unable to be diagnosed. They recently announced results from their study. Blood samples taken from 580 patients without diagnosed cancer returned five positive signals, and of those five, two have been diagnosed with cancer.
Consumer genomics continues to gain traction. Genealogy tests and sensitivity to caffein tests created a 44% growth in services. These tests when desktop or portable sequencers are on tap lend themselves well to the businesses of beauty and spa centres, cosmetology clinics and fitness centres.
LifeBit raised a $3М seed round. They are developing a cloud- and machine learning based software to allow their customers to gain actionable insights into their sequenced data. LifeBit do exactly what an NNU can do with part or full set of sequenced data at the edge by the way. LifeBit and similar companies are from the DNA sequencing as a service sector per se, from that part that should help the service become full-blown and well-rounded. DNA sequencing as a service is akin to how businesses often don’t own their hardware+software but use the compute resources provided by other tech companies as a service. LifeBit, NGX Bio and Meenta – the last two also provide access to sequencing machines – if they pan out, the manufacturers of DNA sequencers will get another effective way to communicate with the market.
As the new batch of important and interesting news appears, so will the next Information letter.