AI- A promise for improving efficiency and accuracy in drug discovery
In a highly competitive business environment, the first company to patent a new chemical entity has all the advantage, and hence, the idea of fast tracking the drug discovery process using artificial intelligence (AI) seems promising. While adoption of AI
Apast few years have been revolutionary for pharmaceutical industry. We have seen a significant transformation in pharmaceutical research and development practice, which is largely driven by increasing importance of patient centricity, value based care and focus on reducing drug development cost.
10 years back, with limited digitalization, data availability was a challenge. However, in today’s world, we see that the healthcare sector is sitting on a data goldmine and consequently we are seeing an uptrend of utilizing big data science in drug discovery and clinical trial activities. As per IBM, healthcare Industry generates 2.5 Quintillion bytes of data everyday in form of patient records, diagnostics records, healthcare records and other scientific databases. However, the important part is—How do we utilize this information? So far we have seen that not every patient is same; not every therapy is behaving the same way for every patient, and hence companies are looking to utilize patient focused insights to develop drug. However, it is almost impossible to handle, manage and analyse this volume of data manually, and that’s where the concept of Artificial Intelligence (AI) comes into picture.
What is artificial intelligence (AI)?
AI is not a new concept, in fact, a large part of theoretical and technological concepts was developed over the past 60 years. However recent advancements
in machine learning and deep learning technologies have developed the practical application aspects and therefore, companies are increasingly looking to utilize AI in their operations.
AI refers to a set of technologies which are converged to sense, understand and act with the ability to learn from experience and adapt over time. AI includes big data analytics, natural language processing (for supporting translation, pattern recognition, visual perception and decision making), and machine learning (for developing computational approaches to automatically make sense of data).
Need for involving AI in drug discovery
The efficiency of drug discovery and trial process— defined as number of drugs approved vs. total R&D budget—has continuously declined over the past few years. The average cost of drug discovery and development comes out to be more than $2.5 billion and it may take more than 10 years for companies to come up with successful candidates. In process of relevant compound identification, scientists and researchers scroll through thousands of research papers, published literature, patents, and disease databases, to understand relationships between biological entities such as genes, symptoms, diseases, proteins, tissues, species and candidate drugs, which is a time consuming process. On top of it, a large chunk of information is being updated on a daily basis. In fact, more than 10,000 medical research papers and literature get published everyday. Considering the effort, it is a humongous task, and despite this, there is no guarantee that this identified and shortlisted compound will be successful in later stages. In fact, 90 per cent of these shortlisted compounds fail to reach the approval stage, and spend on these failure compounds account for about 70–75 per cent of the total drug development cost.
Given this business scenario and increasing cost pressure on pharmaceutical companies, researchers and scientist community is hoping to combine the existing knowledge from previous drug discovery projects with new and existing experimental data, to drive AI-based drug discovery and design process.
AI is expected to streamline and speed-up some key drug discovery activities including identify new drug compounds for screening during the early stages of drug discovery; understand and study new therapeutics applications for previously tested compounds, improve new compound design process and support development of patient-specific therapeutics. Further, an expected higher accuracy from automated systems will likely reduce the late stage failure rate. Considering the strong promises and experimental success in these applications, AI has gained strong traction from pharmaceutical and life sciences IT companies to cash on the potential opportunity. In fact, all the leading pharmaceutical companies, including J&J, Novartis, Merck, Pfizer, Roche, and Astellas, are pursuing AI capability and applications through external partnerships or in-house capability development. As per Frost & Sullivan’s recent estimates, AI-based drug discovery market will exceed $3 billion by 2022.
Life Science IT Start-ups taking lead in the business, with candidate-as-a-service to become reality in pharma industry
As preference towards data-driven decision making in
drug discovery and development grows, we are looking at a scenario, wherein a large part of drug discovery and trial activities will shift from labs to computers to bring the much-needed efficiency. In the nutshell, we are looking at the convergence of data science and medicine science.
When we talk about adopting machine learning and big data analytics in drug discovery, we look at utilizing petabytes of data—sourced from different patient databases, diagnostics records, healthcare records and other scientific databases. About 90 per cent top selling blockbuster medicines only work for 30–50 per cent of the patients, and hence patient centricity is another key trend and challenge, expected to be solved by AI. However one important question which is yet to be answered is—how do we find ways to effectively collect different types of data for better analysis; and more importantly how do we use available information to get the desired results? While strong progresses have been made in this direction, the industry still awaits a permanent and universal solution to this challenge. Another core question is that—is publically available data is good enough to overcome these challenges? Or do we need to do more with data collection? Providing right set of data to AI and machine learning platforms is still a work-in-progress and this is an existing challenge for companies for implementing AI and patient centricity concepts in drug discovery practices.
So far we have seen a number of start-ups venturing in this segment, especially in the US and Europe, and candidate-as-a-service is a concept in development in the industry. In near future, we expect a number of start-ups and research focused companies offering this solution using their in-house AI and machine learning platforms. Accordingly, we expect number of partnerships and M&As activities to increase in the next 2–3 years. Companies, with exclusive access to specific patient and healthcare data, will get higher traction from clients and investors.
Further, with invent and convergence of genomics and next gen-computational chemistry techniques, the scope of applying artificial intelligence in drug discovery will increase, which in turn will result in the development of personalized and precision medicines.
Asia – A slow starter in AI adoption race
Traditionally the convergence of data analytics, AI, machine learning and medicine sciences in Asia Pacific (APAC) market has been slower than the US and European market. The fundamental reason behind this is that most APAC markets are not R&D intensive and the industry revolves around generics and biosimilars. In research focused markets, including Japan and China, availability of databases and language barrier have been key challenges. However, some large Asian pharmaceutical companies are increasingly adopting AI-based drug discovery practices by engaging external vendors. For example, Dainippon Sumitomo Pharma has engaged Exscientia to work on small molecules drugs. Similarly, Astellas Pharma has signed a partnership with big datadriven bioinformatics company NuMedii to conduct drug repurposing projects using machine learning platform. In addition, the Japanese government has also initiated a project in collaboration with some pharmaceutical and technology companies to develop AI-based drug discovery solutions. However most of these projects are in initial phase and we are yet to see a widespread adoption and implementation of machine learning and AI concepts in APAC market.
AI offers a lot of promises for improving efficiency and accuracy in drug discovery. In a highly competitive “winner takes all” business environment, the first company to patent a new chemical entity has all the advantage, and hence, the idea of fast tracking the drug discovery process using AI seems promising. While adoption of artificial intelligence provides a number of business growth and cost saving opportunities for pharmaceutical companies and researcher communities, creating and collaborating with right set of database infrastructure will be critical for the desired success.
TRADITIONALLY THE CONVERGENCE OF DATA ANALYTICS, AI, MACHINE LEARNING AND MEDICINE SCIENCES IN APAC MARKET HAS BEEN SLOWER THAN THE US AND EUROPEAN MARKET. THE FUNDAMENTAL REASON BEHIND THIS IS THAT MOST APAC MARKETS ARE NOT R&D INTENSIVE AND THE INDUSTRY REVOLVES AROUND GENERICS AND BIOSIMILARS. IN RESEARCH FOCUSED MARKETS, INCLUDING JAPAN AND CHINA, AVAILABILITY OF DATABASES AND LANGUAGE BARRIER HAVE BEEN KEY CHALLENGES. HOWEVER, SOME LARGE ASIAN PHARMACEUTICAL COMPANIES ARE INCREASINGLY ADOPTING AI-BASED
DRUG DISCOVERY PRACTICES BY ENGAGING EXTERNAL VENDORS.