Artificial intelligence is not the future, it is already reality!
Martin Pöhlchen, Senior Partner at Alira Health, an institutional bank providing transaction and strategy consultancy, including data management and biostatistics, laid the groundwork for a lively discussion at BIO-Europe Spring® 2019 on how digital technologies are transforming the biopharmaceutical value chain by showing that the digital transformation of healthcare and pharma is already well advanced.
Today, information technology (IT) companies like Apple, IBM and Deutsche Telekom are working together with health insurance firms, such as the German AOK, regulatory authorities including FDA, and big pharma companies such as Merck and Novartis to form new business models and cooperative approaches that directly connect R&D with payers and patients.
Companies which previously had not played an active role in healthcare are now entering the field: Google is today invested in more than 60 biotech startups. Apple fuels personalized care with powerful intuitive tools like iPad, iPhone and Apple Watch generating unimaginably large amounts of live data, directly linked to the physiology of the human body. Facebook pursues its healthcare ambitions by conducting social medical marketing and the “social blood bank.” Amazon provides telemedicine and health apps for existing devices and Microsoft supports promising AI developments with its NExT Projects program—just to name some examples. But all these companies are only laying the foundation for the game-changing companies, startups that come up with new ideas and innovative concepts to change the way we look at healthcare today.
Given all this, Pöhlchen’s first question to the panel was not surprising: Which parts of the biopharma value chain will be impacted most by modern digital technologies? Nora Khaldi, Founder and CSO of Nuritas, believes that all parts of the value chain might benefit from digital technologies in terms of increasing efficiencies and time savings. “The major impact will be on the early stages of drug discovery because the most complex knowledge base is biology itself,” she stated. AI significantly increases the efficiency of research as it shows the most promising ways and higher likelihood of, for example, protein interactions and with this takes away the repetitive necessity and long waits for results.
Bernd Nosse, Global Head of BD&L Technologies, Boehringer Ingelheim International GmbH, sees the quick wins from AI in three main areas: 1) the chance to now also address rare diseases not yet covered given the traditional low success rates; 2) biomarker discoveries that accelerate the development of disease patterns for patient stratification and disease monitoring; and, 3) the significant increase of efficiency in drug development by shortening the timeframe from hit to drug. “AI is not the Holy Grail, but it enhances drug development,” he said.
Sarah Hogan, Partner with Mc Dermott Will & Emery LLP, expects clinical development to benefit most from AI. “There is a lot of data out there on patient population that can help to find the right patients, especially in rare diseases, and to significantly shorten time to market. In addition, public data are considered to help in the approval process and the Food and Drug Administration (FDA) is currently developing a framework for the Real World Evidence Program, a framework that would be used to analyze the viability of using real-world evidence to identify additional indications for drugs already on the market. There is much going on, but it seems the pharmaceutical industry does not fully understand yet how to apply digital data and how to reveal the value of post-marketing data.”
This is not surprising given the traditional world of pharmaceutical R&D is confronted with a very young, very flexible and innovative IT industry. In addition, IT giants look at people (and collected data) in a totally different way—they do not really care about healthy and sick, patient or human being and just collect information, day and night, enabling the seamless observation of the development of a disease. “These real-world data initiatives are great approaches that definitely can help with development of new innovative therapies, but the challenge comes with data integration. These data need to be harmonized, they need to be comparable across geographic, ethnic, or device specific differences and they need to be interpretable. Data will be able to answer any question reliably but if you ask the question the wrong way, the answer will not help you down the road,” said Andreas Posch, Managing Director and CEO of Ares Genetics.“
There are already collaborations ongoing that prove the combination of real-world and clinical data. Last year, Roche acquired Flatiron Health, a market leader in oncology-specific electronic health record (EHR) software with the goal of accelerating industry-wide development and delivery of breakthrough medicines for patients with cancer. “However, you still need the pharmacological insights in a disease to be able to interpret EHR data and look at them in the right context, and the pharmaceutical industry is still not there,” stated Nosse. “There is also the question of data protection and privacy,” Hogan added. “It will be a challenge to build a system that combines privacy, consumers and patients keep asking for, with the need for data retrieval.”
“I believe the biggest task going forward will be to increase the value of the different data sets,” Etienne Bendjebbar, Business Development Associate at OWKIN, said. “There is a biopharma value chain and we are now also starting to see a data value change. Working with raw data taken from e.g., a health app, you can start increasing the value by structuring it in a specific way and in the following have a KOL working on it. You can then use AI to give meaning to the data, you can further take it to an algorithm, to a model and as soon as you are able to create a model you have a potential application.”
But this brings another big hurdle in interpreting data—data management. Most of the (research and clinical) data are published in scientific publications in a format that is not readable by computers. Formats of assays and articles can be very different, the authors use varying technical terms, and the articles are biased by human/subjective opinions. “Editors of such publications should be aware that the next generation of readers are not going to be humans but machines,” explained Khaldi.
AI represents a huge potential for the pharmaceutical industry and despite the fact that there are still some open questions including data privacy policies, IP situation, ownership of data as well as of findings based on AI algorithms and data management, it is already today an indispensable part of the biopharmaceutical industry. AI and pharmaceutical companies together still have a long way to go, but they are well on track to dramatically change the way that drug development goes forward—in the best interests of the patient.