How AI can support drug development
Data-Backed Probability of Technical and Regulatory Success - Here’s How AI Supports Critical Decisions in Drug Development
(Full disclosure: I am working as a part time consultant for Intelligencia and have put together this post together with colleagues of Intelligencia. Thats how I understand this topic 😉 )
Drug development is a slow and risky business, now even more so than in the past. Research shows[1] that the average cost of developing a drug from discovery to market entry is about $2.6B and takes roughly 12 years[2]. You will be hard-pressed to find any other products - except nuclear power plants - with similar timelines and costs.
What makes this even more challenging is the fact that only about 12% of all drugs in clinical trials are actually approved by the FDA[3]. This makes drug development a very long game with odds of success roughly that of a row bet in roulette.
No wonder then, that biopharma companies are looking to decrease development time and cost as well as improve the probability of success. The industry is increasingly looking to AI to help achieve those goals – and bring novel treatments to patients faster.
Identify the Winners Early
Drug development, of course, isn’t roulette.
(Photo by Luka Savcic on Unsplash)
And “winning” or “losing” FDA approval is not based on chance but rather a large number of factors such as efficacy and safety data generated in clinical trials, the mechanism of action or the therapeutic area.
For drug developers, this means that the earlier they can identify promising drug candidates with a higher likelihood to succeed, the sooner they can zero in on development efforts. The reverse is also true. Discontinuing a drug candidate with a low probability of technical and regulatory success (PTRS) early saves potentially hundreds of millions and frees up resources to advance more promising candidates.
Calculating PTRS is an important process at multiple steps in drug development. Making the right decision is especially important in phase transitions where companies need to decide which of a number of drug candidates they take forward or whether to continue (or discontinue) a specific drug development program.
Drug developers have established approaches to evaluate PTRS, such as historical estimates, input from external experts, and statistical analyses, but the low drug approval rates show that there is significant room for improvement. Innovation and new approaches are key to augment current processes to reduce bias and incorporate more data-backed decisions augmented by artificial intelligence (AI). AI should be viewed as a powerful tool in one’s toolbelt that’s capable of improving current PTRS assessments.
AI – the Perfect Tool for the Job
Crunching huge amounts of data and detecting trends and patterns is exactly what AI/ML excels at which makes it the perfect tool for the job and further enhances current approaches to PTRS.
Decades worth of detailed information about many aspects of drug development are available in disparate databases. Information such as clinical trial design, clinical trial outcomes, regulatory data, information about drug biology, and the companies sponsoring the trials can be pulled and curated and then serve as an unbiased foundation on which to base PTRS assessments.
But how can an individual properly synthesize and pull all this disparate data together?
AI is fit for this job which in the simplest of terms are pattern recognition machines. They operate similarly to the human brain in that they learn by “seeing.” That learning enables them to identify patterns and detect connections that were previously hidden. While humans are very good at pattern recognition in their own right, the amount of data that needs to be considered for PTRS assessments far exceeds the human processing power.
One company doing tremendous work when it comes to AI-driven PTRS is Intelligencia AI. Since 2017 this innovative scaleup company has been curating data from a multitude of different data sources as well as developing, selecting, honing and training a variety of advanced AI models with the biomedical expertise needed.
This approach has generated remarkable results. In a retrospective analysis, the company’s AI models accurately predicted 85% of approvals and 81% of failures in Phase 2 trials with no public outcomes.
Industry-wide we are only scratching the surface when it comes to using AI in drug discovery. Strategically applying it to critical decision-making points like phase transitions generates actionable insights and gives drug developers a powerful tool in their attempt to bring life-changing drugs to market more quickly.
Phase transitions are not the only applications for PTRS assessments. This article only touches on the opportunities that AI holds.
Experiences? Comments? What will we see in 2024? Share them below.
If you are at BIO Europe in Munich next week, I look forward to connecting on this topic and others. We are organizing coffee and cocktails on Monday Nov. 6 - let me know if you want to be invited!
[1] “Deloitte pharma study: Drop-off in returns on R&D investments – sharp decline in peak sales per asset”, Deloitte, published January 23. 2023, Accessed October 23, 2023.
[2] PhRMA Research and Development Policy Framework. Published Sept. 2021. Accessed October, 23, 2023.
[3] PhRMA Research and Development Policy Framework. Published Sept. 2021. Accessed October, 23, 2023.