OR WAIT 15 SECS
Marco Anelli is R&D Director of ProductLife Group.
Artificial intelligence offers a number of opportunities in pharmaceutical drug development and manufacturing, but there are barriers to overcome, notably around how well and how quickly the regulatory environment can adapt and keep pace with to the rapid changes.
From discovery and development and through the entire product lifecycle, artificial intelligence (AI) is starting to shake up the life-sciences industry. The potential of intelligent apps (i.e., apps that are capable of learning from experience) is so far-reaching that it is expected that AI-enabled drugs will be brought to market in the near future.
Given the plethora of data that exist across the life-sciences value chain, the ability to gather, manage, and effectively use intelligence from that data has posed a challenge. AI offers the potential to exploit that data in both structured and unstructured forms.
Data that previously were hard to access or too difficult to analyze can now be exploited for a wide variety of purposes, including identifying potential new indications or unmet needs from data collected many years ago. Data leveraged through AI could be invaluable in the approval process, in improving adoption rates, and in gathering real-world outcomes. Moreover, safety and tolerability issues could be identified and handled before they became real problems. Many in the field recognize a strong correlation between AI and personalized medicine, because clear insights into data will make it easier to determine which sets of patients a drug might be best suited to.
According to Christopher Rudolf, CEO of Volv Partners, machine learning is being deployed in regulatory decision making-for example, to determine whether a drug is likely to get authorized or potentially to detect dossier anomalies that might be holding up product approval. Even more impactful is the use of AI in patient identifications when it comes to rare diseases, personalized medicines, the discovery of new indications, or predictions about not only regulatory approval but also reimbursement in markets.
Data in/data out
One of the challenges companies typically face-and an issue that will become exponentially more significant with the Identification of Medicinal Products (IDMP) standards for the identification of medicinal products in a phased program-is the fact that data about products get created, gathered, and stored by multiple different functions in multiple different systems. And accessing that data manually from the summary of product characteristics, clinical reports, and manufacturing reports would require an enormous amount of resources.
AI is considered to have important potential in the IDMP data-gathering process. Gens and Associates found in its 2016 survey, entitled Pursuing World Class Regulatory Information Management, that about half of the 54 companies surveyed say they’re investigating AI for that data-gathering purpose, and another third are monitoring what other companies are doing in that area (1). The data gathered for IDMP can also be fed back into a business for commercial operations-whether for development, manufacturing, supply chain, or marketing.
Elsewhere, AI is already being put to use in the supply chain. In 2016, Zipline, a drone start-up, began dropping blood products in Rwanda by using AI-controlled drones (2). Autonomous delivery, made possible by AI, is likely to become more prevalent across all forms of healthcare delivery by following models adopted by such companies as Amazon.
Regulating for AI
A few pharmaceutical companies are exploring AI in various capacities. GlaxoSmithKline, for example, has been developing AI-enabled apps that provide information for patients, teaming up with IBM’s Watson for a Q&A feature for its cold and flu medication Theraflu (3).
LEO Pharma’s LEO Science & Tech Hub, which was established to build collaborations with life-sciences innovators, is exploring how AI can be applied within drug discovery.
Nevertheless, there are barriers to overcome, perhaps most notably around how well and how quickly the regulatory environment can keep up with AI’s rapid rate of change. Take the development of devices, for example. Self-learning, AI-based devices would continue to learn and adapt, but how the regulatory environment is keeping up with those ongoing changes is uncertain. It could potentially take time-and lobbying by the industry-to get regulatory authorities to adapt guidelines to AI.
As machine learning gathers pace and as its potential to speed things up, improve organizations, and remove cost from every part of an organization becomes clearer, more companies will jump on board and seek a more complete AI strategy.
About the author: Marco Anelli is R&D director at ProductLife Group in Italy and is responsible for the coordination of all clinical and preclinical aspects of projects run internally and on behalf of clients.