From better patient data to better patient outcomes: adding value through a virtuous cycle

Whether through new treatments, better care options or quicker, more accurate diagnoses, bringing #value to patients?is what matters most to everyone working in healthcare. Hospitals provide value through research and care, while pharma and biotech provide value through research, drug pipelines and technology. The value added can be sometimes less obvious for other players, so hopeful innovators must first identify what specific value they are creating for patients – and the tech and investment will follow.?

There is no medical success that does not translate into business success, so successful innovators start with a focus on how the life of the patient is being improved.

By putting the patient first, an entrepreneur creates a virtuous cycle of thinking, from how to integrate the right human insights into an artificial intelligence to asking the right questions in order to reverse engineer AI models into biomarkers and breakthroughs.

Control how you scale

Next comes scale – how can this value be scaled to many patients? This is usually achieved through large distribution channels such as hospitals and partners. Here, control matters. Control of how the value is delivered to patients ensures consistency, but it also allows you to develop a self-improving research feedback loop that will power your next value-enhancing innovation. For example, if a subgroup of treatment super-responders can be discovered in the lab, it could be instantly fed back into real-world treatment decisions in a network of partner hospitals. Owkin has taken this path and is deploying its recently-launched AI diagnostic tools directly to its partnering academic centers.

Financial scale?

Once patient value has been delivered at scale, the focus turns to financial sustainability. Here, all roads lead to the pipeline of pharmaceutical companies, as they provide smarter funding, innovation and value than contract research organizations (CROs). Valuable innovations can be brought in earlier in the value chain in the form of new targets or molecules, or later in the form of the augmentation of clinical trials. Owkin chose to use its federated data network to achieve both.

Choose platform carefully

The primacy of value is clear – the real question is what sort of platform can deliver it, and how much can the platform deliver this value without human intervention?

Make no mistake –?expert knowledge is crucial to any successful platform. In #ai , having doctors (like Owkin’s brilliant Chief Medical Officer Professor Vassili Soumelis) to impart expertise and interpret results is essential. But to deliver value at scale, a platform must be able to add value to a pipeline without a final human decision-maker. It must be data-driven, not intuition or authority driven, which is still the cases in some steps of the drug lifecycle, such as deciding to go from phase 1 to 2, choosing an animal model, a covariate to adjust variance in a phase 3, taking the decision of new combination or sequence of treatment, and many more… As the highest value comes from having a drug pipeline and its associated IP and revenue, many companies turn to experts to create an ‘artificial’ pipeline that ultimately fails, and fails to innovate the pharmaceutical model that is badly in need of innovation. Moreover, most platforms focus on discovery, but the highest value and innovation comes from combining discovery, development and the deployment of diagnosis. At Owkin, we are implementing this into a network of the best academic hospitals in the US and Europe in a data-driven way, without the final approval of an expert.

Owkin’s approach

Horizontally-designed platforms that can inform, augment and accelerate drug discovery, development and diagnosis have the potential to provide both the most innovation and the most value. They span the entire data process, but not just any data –?patient multimodal and highly curated data from the best academic centers worldwide.

Right data matters just as much as the right AI.

We are pioneering a multiscale understanding of this data with causal AI to find impactful multimodal biomarkers. Multimodality matters because the right combination of biomarkers and how they interact changes significantly from one disease to another. While pathology-based models may help to predict a response to immunotherapy in bladder cancer or NSCLC, they are much less effective in DLBCL, in which cells are diffuse and create less focal signals. This approach tackles how these modalities interact, how we can better understand how to link omics to cells, cells to tissues, tissues to pathogenic processes like inflammation, and processes to diseases and outcomes with more causality.

We have pioneered a technology to access this data at scale – #federatedlearning . It offers a proven way to access data at its home, without sharing it, while preserving privacy, and above all while breaking research and competitive siloes to create an unprecedented collective intelligence. The results of MELLODDY, our unprecedented pharma industry AI collaboration , and HealthChain, our equally unprecedented academic centre AI collaboration published last month in Nature Medicine , demonstrated the real-world benefits of federated learning. We access highly-curated multimodal data and samples in the best academic hospitals, curated by the best KOLs. We continuously enrich and curate this data by adding new single cell modalities or digitizing the slides. This is capital intensive – but we are proud that academic centers can freely use Owkin for internal research and innovation. We proudly don’t own any data and reject the idea of a data economy completely – we instead participate in the value economy.

Of course, the right platform needs to be accompanied by powerful engines that can create high value solutions for pharma partners, bringing value to both patients and pipelines. Integrating biomarkers within a drug lifecycle requires a lot of R&D, innovation, and novel methodologies – but workflow integration is just as important as the data and AI. These engines, enriched by expert knowledge (in our case our scientific advisory board led by Professor Miriam Merad and our KOL network of hundreds of top tier medical professors), have to deliver data driven solutions and decisions without human intervention.

The key lesson for us in the AI/medtech sphere is to focus on the value we want to add to patients, with who we can create it and with which platform, and how we can create the right feedback loops to allow the platform to bring patient value in the most data-driven way. From patient to patient – innovation is a near-infinite loop.