According to McKinsey & Company’s overview, technology and innovation in research and development for drugs could speed drug discovery by up to 500 days. Artificial intelligence (AI) can perform critical analysis and work with the large amounts of data that are essential for drug discovery and development. In a field that has grown slower and more costly for years, AI is making a big difference.
Drug Discovery: Time Consuming and Expensive
A 2018 survey of 27 anticancer drug development processes published in the Journal of Cancer Medicine found that it took an average of 6 to 12 years to develop a new anticancer drug from discovery to approval. The survey authors concluded that if this rate was sped up to 5years, over a million years of life could have been saved.
In addition to benefits to patients, McKinsey reports that the top 20 pharmaceutical companies spend about $60 billion a year on drug development. Each drug that is brought to market costs $2.6 billion on average, a cost that has gone up 140 percent over the past decade.
Rapid Advances in AI
Technology is advancing rapidly. Think about the gene editing technology CRISPR and its influence on the understanding of health conditions and drug design. Other technologies are rapidly developing. According to the Washington Post, in August 2022, DeepMind, the AI initiative of Alphabet, provided a snapshot analysis of over 200 million proteins, almost all proteins known on Earth. The resulting technology is so helpful for drug discovery that Jay Bradner, president of the Novartis Institutes for BioMedical Research, said “I’m on it [DeepMind] more than Spotify.”
Some industry experts are comparing DeepMind’s ability to provide snapshots and analyses of any protein to “protein TikTok”. In the future, there may even be a “protein YouTube” based on AI and its ability to analyze and organize huge amounts of information.
A New, Faster Drug Discovery Journey
McKinsey’s report on drug development highlights a framework for excellence in R&D. The goal of the framework is to allow medicines to reach patients faster, improve insights and decision-making, and to reduce drug costs. McKinsey suggests that “drug development organizations will need to create a data strategy alongside their technology strategy” to ensure secure and completely compliant management and access to the large amounts of data they will need. AI can be a crucial tool in establishing a secure, and faster data process in the drug discovery journey.
Other factors in McKinsey’s recommendations for a better, faster new drug development process include:
- Taking a patient-centric view to provide a clear vision of individual drug discovery and the organization as a whole.
- Leveraging innovation from sectors outside of the pharmaceutical industry, including analytical techniques and data management and visualization.
- Adopting cultural change to break down organizational silos, analyze ambiguous data, and move to a more data-driven way of working.
Starting From a Strong Foundation
At early stages in the drug discovery process, decisions can be unclear and discovery processes can be time-consuming. This is whereAI can come in. AI-guided software can help to ‘fill in’ sparse data to inform better decisions and uncover high-quality compounds.
Rather than relying on human analysis, AI can also identify hidden opportunities as well as flag false negatives and outliers, improving decision-making confidence. And, AI can also help to glean insights from prior work to focus on the critical measurements necessary to move forward in the drug discovery process.
An example of a drug discovery tool that harnesses AI in this way is Cerella. Powered by a specialized type of deep learning method, called Alchemite, this AI drug discovery software helps maximize the success of early-stage exploratory work by filling gaps in sparse data (whether it’s missing, uncertain or inaccurate) and suggest novel compounds worth exploring further.
Conventional QSAR modeling can result in decision-making impasses, whereas AI drug discovery software can more accurately predict complex endpoints. Finally, deep learning can provide multiple iterations in modelling which are superior to conventional modelling methods.