Every Compound a Candidate: experience-led risk-taking approaches to accelerate small-molecule drug discovery

Area of Science:

• Drug Discovery
• Computational Chemistry
• Pharmacology

Background:

• Small-molecule drug discovery is a slow and expensive process.
• Artificial intelligence (AI) and machine learning (ML) offer potential solutions but are insufficient alone.
• Current industry timelines for candidate delivery are lengthy, averaging 4.0 years.

Purpose of the Study:

• To highlight the necessity of improved translational predictivity in drug discovery.
• To advocate for a strategic shift towards an "Every Compound a Candidate" approach.
• To reduce drug candidate delivery timelines to under two years.

Main Methods:

• Leveraging AI and ML in drug discovery workflows.
• Optimizing processes and decision-making for faster data turnaround (5-day).
• Analyzing downstream in vitro and in vivo model data for translational insights.

Main Results:

• Achieved a candidate delivery timeline of 2.9 years in a partnered portfolio, compared to industry average.
• Demonstrated the importance of experience-led risk-taking.
• Identified critical blockers and translational thresholds through compound progression analysis.

Conclusions:

• AI/ML are essential but not sufficient for efficient drug discovery.
• An "Every Compound a Candidate" strategy, coupled with improved translational predictivity, can significantly shorten timelines.
• A mindset shift towards proactive learning from all compounds is key to future success.