A: NuMedii, has been pioneering the use of Big Data, artificial intelligence (AI) and systems biology since 2010 to accelerate the discovery of precision therapies to address high unmet medical needs. Artificial Intelligence approaches are a natural fit to harness Big Data as they provide a framework to ‘train’ computers to recognize patterns and sift through vast amounts of new and existing genomic and other biomedical data to unravel diverse complex biological networks involved in disease processes. We use multiple AI approaches, ranging from classical machine learning techniques to newer deep learning systems, to rapidly discover connections between medicines and diseases at a systems level. Our AI approaches are also being used to identify subsets of patients and therapies that are likely to modulate these complex networks for each patient subgroup. As we have worked extensively with Big Data and AI, we have developed a deep appreciation of the limitations of AI. At NuMedii, we prefer to think of AI more as “Augmented Intelligence” than “Artificial Intelligence.” We couple AI, Big Data and systems biology with human intelligence enabling our scientists to have access to more and better synthesized information than otherwise feasible. Our goal is to use this combined system of human + machine intelligence to help speed up drug discovery, cut R&D costs and decrease failure rates in clinical trials, all of which can eventually lead to better, more precise medicines.

A: As a next generation biopharma company, NuMedii has built a powerful technology – AIDD Technology – that harnesses Big Data and AI to rapidly discover connections between drugs and diseases at a systems level. The company has efficiently extracted information from a vast array of disparate data stores to create a structured, proprietary data resource spanning hundreds of diseases and thousands of compounds. NuMedii’s proprietary AI algorithms enable the company to extend well beyond conventional ‘target-centric’ drug discovery approaches by facilitating the exploration of favorable ‘poly-pharmacology’ profiles that can potentially improve therapeutic efficacy by modulating effects on multiple disease pathways.

A: I am the founder and CEO and a molecular biologist by training with more than 17 years of experience turning cutting-edge scientific concepts into products that benefit patients. The potential for AI to significantly alter the landscape in medicine is huge, and there is a lot of excitement and interest in this space. We are seeing sizeable, established companies jumping in with large-scale investments, as well as the launch of hundreds of small startups. Part of the excitement comes from the fact that healthcare comprises a large portion of the U.S. economy, and thus is of interest to many companies, especially tech companies. Today there are AI companies that now touch each of the four key stakeholder pillars: patients (e.g., Apple), providers (e.g., Sense.ly), payers (e.g., Optum, GNS) and pharma (e.g., NuMedii, Benevolent AI, Berg).

A: Artificial Intelligence will continue to have increased impact on multiple aspects of healthcare. Over a period of time, quantum computing is projected to become more readily available, enabling new applications in both the front end of the drug discovery process and also downstream in clinical development. Once we have a few successful examples of how AI has streamlined and enabled us to expedite drug discovery, we will see broader adoption and expect that AI will be routinely used in R&D in the next five-to-ten years.

Most AI start-ups have ended up working with pharma and biotech partners in various ways to help them with their drug discovery efforts. Going forward there will continue to be several providers of AI services. But we will also see more startups become drug developers themselves – a full integration of both AI-driven drug discovery coupled with serious drug development capabilities – to speed up the process all along the way. At NuMedii, we’re certainly going down this road ourselves by developing our own pipeline – we refer to it as “eating our own caviar.” Several others like BenevolentAI, Recursion and Berg Pharma are also developing their own pipelines, and we expect to see more companies follow in our footsteps.

A: Drug discovery and development are highly data-intensive processes, with disparate types of data being generated (from molecules to clinical trial encounters) and a lot of information being tracked. These processes have historically been trial-and-error ridden processes, with high failure rates and costs that are in the billions. Several factors have contributed to these problems: biology is inherently complex and disease manifestation in patients varies across the patient population. Genetic, environmental and other factors also determine how a disease will progress and how patients respond to a given therapy. Thus, these R&D processes and the variabilities involved become a Big Data problem. Artificial Intelligence, coupled with correct data, has the potential to make drug discovery and development less error prone and increase the likelihood of success both in trials and the real-world setting. The hope is that ideally, with AI, some predictability could be regained in these processes.

Compute power is certainly no longer a rate-limiting step for the use of AI in drug discovery. Effective use of AI requires large amounts of relevant, high quality and consistent data to train algorithms for accurate pattern recognition. These data can be particularly challenging to both in terms of access, as well as ensuring that the right data are used for a discovery project. Often data are kept in silos and spread across organizations. In addition, biomedical data are very diverse, spanning multiple “omics” information, such as genetic, genomic, proteomic and metabolomics, as well as environmental exposures, ranging from chemical structures to clinical information. Mapping relationships between these diverse data are challenges that need to be solved in order to make effective use of these data for applications of AI. One of the biggest challenges in this space is a well-quantified cohort dataset that can help train algorithms with a “true” pattern. For instance, if we had high resolution datasets from patients where we collected a multitude of “omics” information and had corresponding longitudinal clinical data, we could then use these datasets to train our AI systems and generate novel insights and accelerate drug discovery.

As of today, there are no successes of AI-driven drugs either approved by the FDA or even in development so companies are reluctant to broadly adopt AI-based approaches in their R&D programs. Once there are a couple of success stories, we should see expansion of such approaches into regular workflows. The good news is that there is strong interest from pharma and biotech to see whether AI can help. We are seeing adoption of AI approaches in discrete parts of the R&D pipeline – from early stages of discovery, to clinical development at smaller scale vis à vis pilot projects. However, we are far from the point where AI can fully automate the drug discovery process and there still is a great need for people skilled in understanding the positive attributes and limitations of AI to make best use of the output of these technologies. And, the lack of people cross-trained in both AI and traditional drug discovery and development is another barrier to broad adoption of AI.

A: I think it’s important to demystify what AI is. Many people interchangeably use many different terms to refer AI such as machine learning, cognitive computing, Big Data or data sciences. Artificial Intelligence or machine learning is a set of software tools that enable us to find patterns in data – either patterns one might be trying to look for or patterns one didn’t know or wasn’t expecting to find. For instance, one could train software to look at measurements from patients who went on to do poorly with a treatment and then use the trained system to look for new patients who might be predicted to do poorly. Identifying patient types early on could be useful as it could help doctors as they consider optimal treatment options, and in turn, could save precious time and possibly extend patients’ lives. We need to think of AI as an important tool in a toolkit – it is only useful if we clearly define the problem we are trying to solve, and the end users are trained to understand both its benefits and limitations.

At the end of the day, the true value of AI to the end user – the patient – is not how we come up with an effective drug but how soon we do so. To that end, AI is a great tool that, if used correctly with the right data sets, can yield revolutionary therapies for the people who need them the most.