Deep Science Ventures Year 3: A new paradigm for Applied Science. Part 2.

Mark Hammond
Jan 31 · 9 min read

Part 2 of 3

You can find Part 1 here and part 3 will be published shortly.

In part 1 we outlined the need for a model that starts not with IP, idea or skill-set, but instead with a much more systematic approach to unpick the changes needed at both a human and technical level across a sector. Some of the most pressing, and challenging, questions are in therapeutics and the urgent need to reduce carbon emissions, so that’s where we started.

Tackling biological complexity in therapeutics

Therapeutics fail for a wide range of reasons including off-target effects, poorly chosen models leading to incorrect assumptions, and lack of effective stratification. The underlying reason is a singular focus on a given target, derived from a point discovery, without sufficient regard for the wider picture. It’s actually rare that the confounding factors causing the failure of a therapeutic were completely unknowable, or that there was just no scientifically feasible way of addressing them. We explored this in more detail in our therapeutics thesis

After 20 years of failure under the Alzheimer's amyloid hypothesis this seemed like a good place to start.

Read the article on Wired

As we recently covered in Wired, one of the most pressing examples of ignoring the complexity of biology is the amyloid story. Anti-amyloid antibodies can be thought of as little hooks that take away broken proteins, the increasing presence of which appears correlated to Alzheimer’s. Drugs like these always work in animal models because the models are specifically designed to express the broken protein, that is the one different thing in those cells and the antibodies clear the problem. Unfortunately it has long looked as if in the real disease there are other errors beyond the presence of the broken proteins.

This reality is largely lost amidst the bias caused by sunk-cost fallacies that drive pharma companies to invest hundreds of millions in late stage trials, based on exceptionally dubious statistics. This is a vicious cycle which has led to nearly 20 years of failed trials and billions of pounds in lost value.

It’s not that other potential approaches don’t exist. For example, the ‘inflammation hypothesis’ has extensive causal data across neurodegenerative diseases. The hypothesis being that it is our own immune system’s over response to cell damage that drives the loss of neurons vs. the gradual loss of function of internal cellular systems. The absence of a manufacturable therapeutic for something like the inflammation hypothesis is because it falls outside of existing research trajectories in academia, which is in turn because such work is comparatively less likely to lead to Nature papers or Nobel Prizes. This is a clear example of the massive systemic effects of our sole reliance on tech transfer from academic research, which indexes primarily to novelty and discovery.

To work towards building a therapeutic that solves this challenge, we brought in Dr Timothy Newton, an expert in neuron proteostasis to join us from the Alzheimer’s Research UK Cambridge Drug Discovery Institute. Borrowing from the fields of oncology and synthetic biology we brought together a team of renowned inflammation regulation and modelling academics including Dr Adrian Liston at Cambridge and Dr Andrea Serio at Kings. We have produced early evidence in unbiased models and are initially focusing on ALS as a more rapid route to market as it shares many of the same features as Alzhiemer’s and Parkinsons but sadly is far more aggressive, this provides both the opportunity to treat this disease in addition to Alzhiemer’s and enables shorter, clearer trials.

Tim Newton, CEO of Reflection Tx speaking at the FT Global Pharma Conference

What if immunotherapies could be as simple to produce as gene therapies?

Another example of the strange contrast between incredible science and less than ideal application is that of immunotherapies. This approach of re-training immune cells to spot cancer has shown incredible results in patients. However, its effectiveness and patient applicability is limited by the highly artificial nature of culturing cells outside the body. This begs the question, given that gene therapies have shown efficacy in patients and our bodies are a cell production factory: why can’t we modify immune cells directly in the body?

Bakul and George, founders of ImmTune.

We brought on Dr Bakul Gupta and Dr George Tetly, Imperial College nano-medicine and Cambridge immunology postdocs respectively to investigate what is holding this back. We quickly found that in the main it was down to biases and outdated beliefs rather than insurmountable technical constraints. We set about designing a plan involving three distinct areas of research across three continents. Due to the competitiveness of the field the company, ImmTune, and associated academics will remain in stealth for the next 6–12 months.

This adds to our previous work in oncology including ConcR’s single cell level resistance prediction, in which they’ve already demonstrated that they can predict the outcome of a given drug sequence with over 80% accuracy, potentially saving over a third of lives currently lost to cancer resistance. At the same time, Cytera has completely automated the cell culture required for oncology R&D (and beyond), finally freeing scientists to be, well, scientists. And finally Antiverse, backed by Abcam founder Jonathan Milner, is designing antibodies for oncology targets in-silico which are typically impossible to generate using traditional phage-display approaches.

We will be extending our oncology focus further in the new year, in line with our recently announced partnership with the world’s largest cancer charity, Cancer Research UK, giving our teams access to the most powerful network of scientists and labs working on cancer, as well as closer integration with specialised seed and series A investment.

Are microbiome therapeutics all hype or are we missing something?

Another area that seems to be taking an overly simplistic approach in a complex environment is that of microbiome. Barely a week goes by without the microbiome being associated with another disease, ranging from IBD and Parkinsons to Cancer. Yet the only effects that have so far played out in clinic is full replacement of the gut microbiota via fecal transplants (with 90% success on extreme cases of C.diff but effective in 0% of patients with IBD) and the impact of bacteria on check-point inhibitors in cancer. Indeed even companies backed by leading funds have committed fraud to try to make microbiome diagnostics seem like a real thing.

Stock photo of phages attacking bacteria because tiny error bars don’t cut it in blog posts. Source: AMI Images/Science

We wanted to ask why this is and brought in Harvard biochemist Dr Alexandra Sakatos and Manchester synthetic biologist Dr Matt Cummings to explore. We quickly realised that nearly every microbiome therapeutic considers only one half of the ecosystem, assuming that bacteria is the main driver of disease and largely ignoring the influence of the viral / phage component. We ended up building two distinct but complementary companies in this field. One, Ancilia, in partnership with University of Washington, Harvard and working with the inventors of CRISPR, quickly demonstrated the causal nature of the virome in IBD and designed CRISPR-based therapeutic strategies to restore natural immunity.

Our second area of focus was the problem that in many cases wiping out a given ‘bad’ bacteria doesn’t benefit the patient, it simply drives resistance to the therapeutic used. This includes very large disease groups such as acne and the metabolic effects that drive the differential patient response in most drugs. Rather, a much more effective therapeutic can be built by selectively sensing and modifying levels of a given bacteria. CC Bio have built unique sensing technology to achieve this and are currently finalising pre-clinical work in partnership with a major pharma company.

What’s the highest leverage change we can make to avoid climate catastrophe?

Within our Net-zero energy transition thesis, our focus is on radically new approaches to scale emissions-preventing or emissions-mitigating technologies. We’re focused on creating companies which do not need to reach the scale of an oil and gas supermajor before becoming economically viable, and/or which take advantage of arbitrage opportunities in broken markets, which deploy intelligence and engineered biology to disrupt traditional petrochemicals, and which in doing so, valorise our most abundant resources in scalable, sustainable ways.

To deliver this we scaled up a partnership with the Oil and Gas Technology Council (OGTC), a coalition of most of the world’s most powerful energy companies, to accelerate the net zero transition to support the companies we create, giving teams access to an automatic non-dilutive grant of £100k and intimate connections to the largest energy companies in the world, test facilities and industrial pilot financing.

Batteries are the lynchpin of a cleaner future so why do new battery chemistries so rarely reach the market?

To better understand this, we brought on Nuno Pereira, a mechanical engineer who at the time had term sheets for investment into his existing lithium ion battery company. After nearly 100 interviews with battery company management teams and academics it became clear that the only thing holding back battery technology was that companies are all based on an academic finding, which itself is an incredibly slow search within an enormous combinatorial space.

Early prototype of Holy Grail’s autonomous discovery platform

Battery chemistry development assumes that you can tweak and substitute one component of a battery system (say, the anode) and leave the rest of the system unchanged, then simply scale up. But this doesn’t work in anything related to chemicals or materials, where processes differ radically at different scales. Rather it is critical to solve the combinatorial ‘rest of the battery’ problem in step with designing new materials. To achieve this we built out a venture, Holy Grail, to deploy an autonomous Bayesian learning system and complementary robotics to rapidly identify the global optima in the incredibly large material design space. This system enables research in batteries and other materials to progress up to 10,000 times faster. The team is Advised by Prof. Peter Littlewood, Chairman of the Faraday Institute, formerly senior management at Bell Labs and the team have recently graduated YCombinator S19.

How do we create a sustainable transition from hydrocarbons to renewables?

Developing and enabling clean technology is critical, but realistically the energy majors are by far the largest sources of CO2, are heavily subsidised and aren’t going to disappear overnight. We need solutions that work with the energy industry to accelerate the transition.

Optic Earth is the definition of an energy transition company. They are moving from working initially to identify sources of transition fuels such as natural gas, to identifying ideal sites for geothermal and carbon capture and storage. They do this by massively accelerating our understanding of the subsurface of the Earth. The current process requires the arduous interpretation of seismic (vibration) data, which today can take longer than 18 months, but worse, result in huge inaccuracies. Optic Earth is a team of geophysicists with expertise in machine learning, and within less than a year, they have built automated, rapid software that produces outcomes of the same standard as the state of the art used by the most advanced energy companies. In just 6 months they have reached terms on a seven figure work package with one of the largest energy companies in the world.

Founding Analyst Gael is working on carbon capture technology that works for industry and economically.

Finally, one of many areas that we’re just beginning to explore in energy is that of financial incentives. The typical local optima venture here is software-as-a-service for companies to manage carbon reporting. It’s important, easy to get ‘tech venture metrics’ quickly and raise money, and packed with competition, but it isn’t the full picture. We know that the providers of major flows of capital are beginning to consider the impact of their investments on climate and price that into the cost of capital. Yet the array of schemes, reporting and lack of accounting for externalities are almost impossible to act on. For example, oil companies are actually nowhere near the top of carbon output in terms of self-reported business emissions alone. How can we bridge this gap between capital and carbon mitigation? We think we can see a way forward: if you’re passionate about this space, you can find out more about current opportunities here.

In the final part 3 of this series we’ll cover how we are expanding this platform into new verticals and back along the research pipeline.

Deep Science Ventures

A new paradigm for applied science. We bring together teams of scientists to seize crucial opportunities and redefine industries.

Mark Hammond

Written by

Founder at @deepsciventures creating a new paradigm for applied science. Ex-neuropharmacologist & AI researcher.

Deep Science Ventures

A new paradigm for applied science. We bring together teams of scientists to seize crucial opportunities and redefine industries.

Welcome to a place where words matter. On Medium, smart voices and original ideas take center stage - with no ads in sight. Watch
Follow all the topics you care about, and we’ll deliver the best stories for you to your homepage and inbox. Explore
Get unlimited access to the best stories on Medium — and support writers while you’re at it. Just $5/month. Upgrade