How I use AI to turn failed drugs into new medicines

2026-06-10 · Source: Machine learning : nature.com subject feeds · Field: Science & Research — Life Sciences & Biology, Health & Medical Research, Artificial Intelligence & Machine Learning · Depth: Intermediate, medium

Summary

Ignota Labs, a Cambridge, UK-based company founded in 2021 by Layla Hosseini-Gerami, Jordan Lane, and Sam Windsor, utilizes an AI-driven platform to re-engineer failed drugs. The company secured a US\$6.9-million funding deal in February 2025 and focuses on identifying why drugs fail in clinical trials, particularly due to safety concerns like toxicity. Their platform narrows down thousands of failed drugs to the most promising candidates, then applies deep learning, bioinformatics, chemoinformatics, and multimodal data to understand the root cause of safety issues. The goal is to make minimal chemical changes to these drugs and efficiently return them to clinical trials, differing from drug repurposing by creating new intellectual property. Hosseini-Gerami, the chief data-science officer, combines her chemistry and bioinformatics background to lead the R&D of AI tools, including the SAFEPATH platform, which addresses complex data ranging from molecular structure to living organism responses. Her work earned her a spot on Forbes magazine's '30 under 30' list in April 2025.

Key takeaway

For AI Scientists and Research Scientists developing new therapies, this approach highlights the value of applying AI to rescue failed drug candidates rather than starting anew. You should consider integrating multimodal data and deep learning to diagnose and mitigate specific safety issues like off-target binding or pharmacokinetics problems. This strategy can significantly reduce the financial and ethical costs associated with drug development by efficiently returning promising compounds to the clinic.

Key insights

AI can identify and mitigate toxicity issues in failed drug candidates, accelerating their return to clinic.

Principles

• Combine chemistry, biology, and AI.
• Focus AI on specific biological questions.
• Address drug safety early.

Method

Ignota Labs' AI platform first identifies promising failed drugs with safety issues, then uses deep learning, bioinformatics, chemoinformatics, and multimodal data to diagnose the root cause and develop mitigation strategies.

In practice

• Analyze p*K*a to predict molecular properties.
• Identify biological targets driving therapeutic effects.
• Use multimodal data for drug rescue.

Topics

• AI Drug Discovery
• Drug Re-engineering
• Chemoinformatics
• Bioinformatics
• Deep Learning
• Toxicity Prediction

Best for: Investor, Entrepreneur, AI Scientist, Research Scientist, Director of AI/ML

Related on AIssential

• AI is reengineering drug discovery by speeding up testing and scanning petabytes of data for connections between diseases — AI significantly enhances drug discovery by analyzing vast data to predict targets, efficacy, and disease interrelationships, improving success rates.
• 🔬 Training Transformers to solve 95% failure rate of Cancer Trials — Ron Alfa & Daniel Bear, Noetik — Improving cancer drug success requires advanced AI models for precise patient selection based on multimodal biological data.
• DeepER-Med: Advancing Deep Evidence-Based Research in Medicine Through Agentic AI — DeepER-Med enhances medical AI trustworthiness via explicit evidence appraisal and agentic collaboration.
• 3 Questions: Using AI to accelerate the discovery and design of therapeutic drugs — AI, particularly deep learning and generative models, significantly accelerates novel antibiotic discovery and design.
• Two AI-based science assistants succeed with drug-retargeting tasks — AI-powered agentic systems can accelerate scientific discovery by synthesizing vast literature and generating testable hypotheses.
• Why Pharma’s AI Bet Might Be Wrong | Martin Shkreli — AI's role in drug discovery is limited, with idea generation and human testing as major bottlenecks.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Machine learning : nature.com subject feeds.