For decades, outcomes for cancer patients have improved steadily but slowly. Breakthroughs in R&D were achieved one gene and one protein at a time, and cancer diagnosis often came too late for the patient.
In the 1980s and 1990s, new cancer medicines had been arriving at the rate of only one or two per year. Oncologists generally treated patients with the same standardized approach — a combination of surgery, chemotherapy and radiotherapy, with treatment outcome varying drastically between patients. This is because cancer is a complex disease; cancerous cells from different patients or even the same patient can be very different over time.
Since then, scientific and technological advances have pushed the field rapidly forward. We’ve seen improvements in both diagnosis and treatment. Five-year survival rate of all cancers in the US increased from 50% in 1970s to 67% in 2013¹. Prostate cancer saw the greatest improvement in five-year survival rate, from 67.8% to 98.6%².
DNA recombinant technology and genome sequencing are among the many technologies that have driven the advances. The past few years have also seen an explosion of innovations in AI, immuno-oncology and gene editing such as CRISPR — though with varying degrees of adoption in cancer management.
We at Luminous have spent a lot of time thinking about the next phase of cancer research, drug discovery, diagnosis and treatment. We think that the push from data analytics and automation will be particularly strong and far-reaching, bringing impact to all stages and all players involved.
We thought it would be useful to give a recap of the progress so far and discuss how software, robotics and data can help us march forward in the fight against cancer.
Targeted therapy — the successful first steps
Since the 1980s, it has gradually become clear that there is no “one-size-fits-all” approach for cancer. Now the world is advancing towards precision medicine, i.e. using information of the patient’s genes, lifestyle and environment for prevention, diagnosis and treatment. Targeted therapy that acts on specific proteins on the surface of cancer cells is the cornerstone of cancer precision medicine. A successful example is trastuzumab (Herceptin), developed by Genentech for breast cancer. Approved in 1998, Herceptin is still a leading breast cancer drug with annual sales ofaround $7 billion, making it the 5th highest selling drug currently³. Targeted therapies have limitations, however, as the constantly evolving nature of cancer cells can lead to resistance which can cause targeted therapy to fail.
Immuno-Oncology — our strongest new tools
The field of immuno-oncology has since taken off. While there are currently no more than ten approved drugs in this space, close to 4,000 candidates are in the pipeline. Immuno-oncology is about enabling the immune system to kill cancer for us. While the first immune-oncology drugs approved for the market are checkpoint inhibitors — the top selling ones being Bristol-Myers Squibb’s Opdivo and Merck’s Keytruda, the frontier of immune-oncology is CAR (chimeric antigen receptor) T-cell therapy. An advantage of CAR-T cell therapy is its long-lasting effect. CAR-T therapy is a “living drug” — since the T-cells persist in the body for long term, they keep recognizing and attacking cancer cells, preventing disease relapse.
But what about software and data?
Software for life sciences
A major challenge in advancing cell therapies lies in manufacturing and scaling up. As CAR-T therapy mostly requires modifying the patient’s own cells, the production is extremely complex and labour-intensive. They are also prohibitively expensive — Kymriah’s list price is $475,000 for one course of treatment! The key to making these therapies accessible lies in robust, scalable and industrialized manufacturing processes, which in turn require sound quality control, high level of automation, and deep and precise understanding of the process intermediates and the final product. This is not possible without superior data analytics — or better, software that can analyse the complex data intelligently and present new insights about the processes. A useful market overview of the role of frontier technology in cell and gene therapy can be found here.
Besides manufacturing challenges, the field of life sciences R&D in general can also benefit from this computer-aided approach. The returns from drug discovery have been declining over the past decade, with the term “Eroom’s law” (Moore’s law backwards) being introduced to describe this decline in R&D efficiency. While many factors have contributed to the reduction in productivity, the “basic research-brute force” bias is intrinsic to how experiments are commonly conducted in the lab. This calls for a new methodology in life sciences research.
Luminous Ventures has invested in Synthace, a London company whose digital automation and statistical experimentation technology helps optimize complex biological processes and systems and thus increases productivity in life sciences. Its software platform has been applied to gene therapy and bioprocessing. Case studies have seen a 70%-80% time saving and up to 30% resource saving.
Similarly, Utah-based Recursion Pharmaceuticals combines computer vision with high throughput and high content microscopy to enable phenotypic drug discovery at scale. This enables them to analyse images from millions of experiments to identify promising drug targets. This seemingly intuitive “full stack” method is in fact radically different from the manual approach that traditional target-based drug discovery depends upon. Recursion Pharmaceuticals raised $121 million in July 2019.
Insitro (a play on in-silico and in-vitro) is also adopting an approach analogous to that of Recursion but looking to create novel large data sets validated by a tight closed loop of in silico and in-vitro cell-based experiments. Insitro has raised $100 million from investors including Andreessen Horowitz, GV, Foresite Capital, Arch Venture Partners, and Third Rock Ventures.
A useful overview of software platforms using AI for drug discovery can be found here.
Automating early detection and treatment
Creating amazing therapies is only one way of beating cancer. The best approach is catching the disease early, since early detection of cancer greatly increases the chances for successful treatment. More than 90% of women diagnosed with breast cancer at the earliest stage survive for at least five years, compared to only 15% for women diagnosed with the most advanced stage of the disease.
We can now diagnose cancer earlier, thanks to the developments in AI and computer vision in imaging and biopsy. Deep learning of medical images has been undertaken by researchers and companies for cancer detection and diagnosis. In academia, a recent study has shown that AI can outperform radiologists at lung cancer screening. A team at MIT and MGH have demonstrated that their deep learning model can recognize subtle patterns in breast tissue that are precursors to malignancy, therefore able to predict if a patient is likely to develop breast cancer as much as five years in the future.
There are numerous AI-powered radiology imaging startups. Besides having access to diverse datasets, successful start-ups must also integrate into existing healthcare workflow by demonstrating value for clearly defined use cases, as well as helping healthcare practitioners manage workload and improve accuracy. Luminous has invested in Optellum, an Oxford-based company developing AI clinical decision support software for lung cancer management. Optellum’s software will enable clinicians to diagnose cancerous lesions more accurately, ensuring timely intervention and also avoiding needless surgeries and biopsies where there is no risk, saving billions for hospitals and payors.
Early diagnostics aside, there is also immense promise for robotic surgical devices in cancer treatment. In early 2019, Johnson & Johnson announced its acquisition of Auris Health for $3.4 billion, with an additional $2.35 billion payable upon reaching agreed milestones. Auris Health has developed a less invasive and more accurate robotic approach for cancer surgery. Their FDA-approved platform inserts a flexible robot inside the human body and allows doctors to navigate inside.
We think these only mark the beginning of using computers and machines for early diagnosis of cancer and surgery. The trend will continue.
Systems that make sense of real-world data
The final string to the bow is real-world evidence (RWE), which the FDA defines as “Healthcare information derived from multiple sources outside of typical clinical research settings, including electronic medical records (EMRs), claims and billing data, product and disease registries, and data gathered by personal devices and health applications.”
As evidence for the growing activity and interest in the area, Roche acquired Flatiron Health for $1.9 billion. Flatiron’s database platform organises oncology-focused information and makes it useful for patients, physicians, and researchers. The acquisition represents a step closer to precision medicine driven by real-world data.
Healthcare is transitioning to an era where increasing focus is placed on outcomes and value, and RWE is a key part of this. Traditional medical interventions based on sporadic or periodic interactions between doctors and patients are shifting towards a holistic approach where all relevant data can inform care decisions in real-time.
This is made possible by our improved understanding of the external factors and the growing prevalence of health-related data, including data from clinical trials, electronic health records, genomics data, data gathered by wearable sensors and more.
Stakeholders across the healthcare system can all benefit from real world data for making decisions. Pharma companies have integrated RWE in their end-to-end product cycle to advance disease understanding and clinical guidelines, and support regulatory and outcome-based reimbursement decisions.
On a more granular level, RWE can be valuable for clinical study design such as identifying eligible participants and patient recruitment. It can be used as synthetic control arms, reducing time and cost of clinical trials. When it comes to reimbursement, healthcare providers are increasingly relying on RWE for outcome-based pricing of therapies. This is particularly relevant as expensive gene and cell therapies will continue to appear. As an example, Spark Therapeutics, the developer of the gene therapy Luxturna — its list price being $850,000 — agreed with the health insurer Harvard Pilgrim that if improvements in patients are not seen by the 30-month mark, Harvard Pilgrim will receive a rebate. Clearly, pharma companies, regulators and health insurers need to work together to figure out how to pay for these therapies, and the analysis of RWE can help.
The use of RWE has just started. Significant data challenges need to be conquered. Limited data access, uneven quality of data sources, and the lack of robust and standardized RWE analytics are some of the key barriers. The startups that can solve these challenges are bound to deliver tremendous value. Flatiron’s platform can organize structured and unstructured oncology data from diverse source systems; based on these data, its analytics engine can pull out relevant insights, which help to stratify patients for clinical trials, improving trial results or enable innovations in R&D.
In both the US and Europe, we have seen deals that signify increasing interest and activity in this space. Aetion in New York has developed a platform designed to offer real-world analytics and evidence to life science companies and payers. Founded in 2013, the company has raised $77 million to date from investors including Sanofi, UCB Pharma and McKesson Ventures. Aetion seeks to inform therapy development and improve outcomes by evaluating real-world data from claims, EMR, registries and clinical trials.
In Europe, Oxford-based Sensyne Health IPO’ed AIM in 2018. The company seeks to create value and accelerate drug development through the analysis of anonymised patient data in collaboration with NHS trusts.
Just a few weeks ago, Medidata, a US SaaS company in healthcare data was acquired by Dassault Systèmes for $5.8 billion.
The old-world view of getting sick, speaking to a doctor, procuring a diagnosis and then moving onto the treatment of a disease which is already very late needs to change. Further, the cost of R&D and the development of new therapies needs to come down. Software, robotics and data are a key part of the solution — these can drive earlier stage diagnosis, form key understanding of risk factors for individuals, cut the costs of development of new targeted therapies and robotics has the potential to increase surgical outcomes. There is a lot to be done but, reassuringly, we live in a time where all these advances are within reach. Luminous Ventures is looking forward to this future and is working hard to make this happen.