Screening Synthetic Biology Companies
I asked a person on my team at SC Capital to help me explore an idea using his background as a PhD in economics and statistics. After a lengthy discussion about biological data sets that have the quality of non-mean reversion, he resisted my concept and instead suggested that data are almost always a merry-go-round. I took that to mean that financial data have the quality of being in motion but almost never go anywhere important. Most financial and economic patterns are noise, and most of the valuable signals have been arbitraged away, at least in the short term (less than a year).
That comment set me on a real safari as I knew I was on to something. Without realizing it, my team member summed up my understanding of the state of small molecule drug development in oncology. Consider some facts about drug development. 72 cancer therapies approved between 2002 and 2014 only bought patients an extra 2.1 months of life compared with older drugs, researchers have found. And there’s no evidence that two-thirds of the drugs approved in the recent years improve survival at all. The life extension property of these drugs has to be further qualified by the quality-of-life question that results when side effects tend to be toxic. Yet the system continues to spin out new drugs that are billions of dollars in the making and a decade in development – with a 5% success rate. The name for this condition is Eroom’s law (Moore’s law spelled backwards). It is the concept that drug discovery is becoming slower and more expensive, despite improvements in technology, a trend first observed in the 1980s. It appears to have improved only slightly in recent years with some new technologies I’m discussing here.
As a non-scientist, I want to develop a screening framework for early-stage companies iterating with a cluster of technologies that include artificial intelligence (AI), machine learning (ML), Bayesian statistics, next-generation pan-genomic sequencing, statistical genetics, computational biology, and quantitative genetics. All of these tools that are used individually – or together in novel ways – can be classified under the umbrella of synthetic biology, and they are reversing the negative trend in Eroom’s law. For me, this means gene and cell-based therapies that create precise and sophisticated responses to disease by reprograming cell behavior and activating immunotherapeutic responses in the world of oncology. What could be the harm in imagining better health outcomes using new technologies where the backdrop is a national healthcare system that today cannot make cancer screening strategies that are risk-based and instead relies on age-based guessing.
The starting point is to understand that these companies cannot be more than six or seven years old because of the recency of innovation in the tools required to do the work. Also, the value created is intangible (and uncertain and risky) and is therefore undiscoverable in the financial statements of the companies pursuing the discoveries.
One thing is certain: the pace of innovation is narrowing the time between scientific curiosity and engineered scaling. Data science (data sets + AI, ML) is creating and releasing new information at a geometric rate. The units of value/compute power yield much more today than in years past, especially with deep learning (DL) and deep reinforcement learning (DRL), a subset of AI that are biologically-inspired neural networks. This yield is far from equilibrium and therefore above-average value creation can persist for far longer than historical experience. Said another way, the pace of knowledge creation trends up without mean reverting, which leads to more inputs to the pace of iteration and knowledge discovery. Eventually, real world impacts emerge. mRNA vaccines come to mind. So does Google’s deep-learning program for determining the 3D shapes of proteins.
One example of a company in the mRNA space is Strand Therapeutics. Its mRNA programming technology promises to make mRNA therapies safer and more effective by programming the location, timing, and intensity of therapeutic protein expression inside a patient’s body using mRNA-encoded logic circuits. Another example is Incitro, an early stage company exploiting recent advancements in cell biology and bioengineering.
In the past, drug development approached the natural phenomena of disease with imperfect information and very slow methods of hypothesis falsification by manual testing. Today, data has become richer and more relevant as the predictive power of machines improves therapeutic yield – and the methods of falsification by testing have greatly improved and have become a data set in and of themselves. Drop the idea that medicine is an exclusive human task and realize that traditional statistical methods cannot discern the complexities of many diseases. Instead, we’re headed towards deep reinforced learning AI that operates at, “unprecedented accuracy, which is even higher than that of general statistical applications in oncology,” according to a review published in the journal Cancer Letters.
I recently interviewed Jo Bhakdi, CEO of Quantgene, in a fireside chat with several hundred attendees on the Clubhouse platform. Quantgene is innovating in the liquid biopsy space. His company is rolling out a direct-to-consumer diagnostic tool that can detect the presence of cancer at an early stage, likely prior to metastasis. Advanced genetic sequencing and AI allow Quantgene to detect cell-free cancer DNA at an early stage from a single blood draw that produces about 10 billion data points per customer. The tests have validated single molecule precision. The company’s algorithm has learned to separate the signal from the noise, and, with use, it improves both the accuracy of the genetic sequencing screening tool and the range of cancers it can detect. Each customer contributes to the value of Quantgene’s data set. Test results are confirmed by traditional diagnostic methods towards a diagnosis that is mediated through a Quantgene physician who has special training in interpreting the data produced by Quantgene’s tests. The company develops a total picture of health that combines liquid biopsy and genetics. I look forward to being a customer, not a patient. In terms of investing in companies like Quantgene, one new factor is understanding a company’s potential value to their customers, then society – followed by evaluating how that value translates into returns for risk-taking investors. Imagine Facebook’s network effects accruing to (predictive) early cancer detection and cures.
Another layer of consideration is the risk posture of participants (scientists/investors/physicians) who may not have considered how new technologies are evolving closer to the workings of natural biological systems (exploiting laws of physics/quantum physics, genetics) and therefore better outcomes per unit of input. Many don’t get it – and indeed this requires a leap – and those same people are often biased by their own risk aversion. In the past, participants assumed a close correspondence between therapy and disease, and the gap in this equation is toxicity. In the future, the two (therapy and disease) will have perfect fidelity because they developed along the same lines. The tricks cancer uses to evade detection of our immune system evolved through iterations that extend over thousands of years. I’m not aware of any cancers that evolved to secrete toxic compounds like the ones we use to defeat cancer. Instead, cancer uses immune editing tools that include immunosuppressant molecules. We’re gearing up to use the same kind of tricks. The era of senseless molecular violence is coming to an end.
Conjuring a bio-health system capable of innovating therapies using the same methods as disease may start without human insight and may have no domain knowledge beyond the basic rules. For example, a DeepMind blog post concluded that AlphaGo Zero used, “a novel form of reinforcement learning, in which AlphaGo Zero becomes its own teacher. The system starts off with a neural network that knows nothing about the game of Go. It then plays games against itself, by combining this neural network with a powerful search algorithm. As it plays, the neural network is tuned and updated to predict moves, as well as the eventual winner of the games.” After millions of games of self-play, “…the system progressively learned the game of Go from scratch, accumulating thousands of years of human knowledge during a period of just a few days. AlphaGo Zero also discovered new knowledge, developing unconventional strategies and creative new moves.” What humans perceive to be random and complex may not be the case with DL and DRL systems. Try telling a scientist that unlocking the mysteries of biology could mean discerning nonlinear correlations and scale variances that do not make smooth transitions from gravity effects to molecular effects to quantum effects.
Dr. Barbara Engelhardt is a scientist at Princeton University. Her work involves combing through vast pools of genetic variation data, and even discarded data, looking for hidden gems. In research published in 2017, for example, she employed a black box model to determine how mutations relate to the regulation of genes on other chromosomes in 44 human tissues. One finding points to a potential genetic target for thyroid cancer therapies. And her work has linked mutations and gene expression to specific features found in pathology images.
Biology on the scale of millions of years is nothing more than a giant statistical computing machine designed to confer evolutionary advantage by twisting the evolutionary process. Data science in the cloud for purposes of curing human disease is the digital extension of human immunity, a quickening of evolution. Every human that contributes to this data in the cloud serves to increase the tools’ therapeutic value. This is a natural form of innovation as it exhibits increasing returns to scale and non-linear outcomes. The opposite of what I’ve described here is the U.S. healthcare system, which is the closest most Americans will come to experiencing the Soviet Union. It is a $4 trillion annual spending machine designed to link clinical symptoms to billing codes, mediated by campaign contributions. It is like Eroom’s law but in human life years.
Consider Repare Therapeutics, Inc, a Canada-based precision oncology company targeting specific vulnerabilities of tumors in genetically defined patient populations. Its approach integrates discoveries from several fields of cell biology including deoxyribonucleic acid (DNA) repair and synthetic lethality. Its SNIPRx platform combines CRISPR-enabled gene editing target discovery method with protein crystallography, computational biology and clinical informatics.
Today, there is a much larger surface area for discovering synthetic biology opportunities and also a large group of people blind to the opportunities. Historical experience/data is of far less value than before, so the best opportunities are with the people/technologies/business models that have the quality of discovery, youth, (positive) rebellion and sharp minds combined with the tools that leverage physics/AI/ML/genetic sequencing. Risk management means spreading bets around to multiple industries, stages of development and therapeutic tools. Public markets are the ideal place to manage risk in these speculative new companies that have the potential for VC-like returns.
Investors are advised to conduct their own independent research into individual stocks before making a purchase decision. In addition, investors are advised that past stock performance is not indicative of future price action.
You should be aware of the risks involved in stock investing, and you use the material contained herein at your own risk. Neither SIMONSCHASE.CO nor any of its contributors are responsible for any errors or omissions which may have occurred. The analysis, ratings, and/or recommendations made on this site do not provide, imply, or otherwise constitute a guarantee of performance.
SIMONSCHASE.CO posts may contain financial reports and economic analysis that embody a unique view of trends and opportunities. Accuracy and completeness cannot be guaranteed. Investors should be aware of the risks involved in stock investments and the possibility of financial loss. It should not be assumed that future results will be profitable or will equal past performance, real, indicated or implied.
The material on this website are provided for information purpose only. SIMONSCHASE.CO does not accept liability for your use of the website. The website is provided on an “as is” and “as available” basis, without any representations, warranties or conditions of any kind.