Source: The use of AI in drug development (The Hindu, May 17, 2024)
Developing new drugs is a slow and expensive journey. But a revolution is brewing! Artificial intelligence (AI) is transforming drug discovery, offering the potential to identify better targets and design more effective treatments, all at a fraction of the time and cost. This raises exciting possibilities, but also important questions.
Accelerating Drug Development with Artificial Intelligence
- Drug development is expensive and time-consuming.
- Artificial Intelligence (AI) can speed up this process.
Initial Steps in Drug Development
Identify and Validate Target:
- Target: a biological molecule (usually a gene or protein) that a drug binds to.
- Most targets are proteins.
- Only proteins with ideal docking sites for drugs are “druggable” proteins.
Discovery Phase
Target Protein Identification:
- Target protein sequence is input into a computer.
- Computer searches a library of small molecules for the best-fitting drug.
- Assumes known structures of target protein and drug.
- Uses models if structures are unknown.
Advantages:
- Avoids time-consuming lab experiments.
- Saves on expensive chemicals and reagents.
- Reduces high failure rates.
Pre-Clinical Phase
Testing Potential Drugs:
- Drug candidates are tested outside a biological system.
- Uses cells and animals.
- Assesses drug safety and toxicity.
Clinical Phase
Human Testing:
- Initial tests on a small number of patients.
- Further tests on more patients.
- Evaluates drug efficacy and safety.
Final Stages
- Drug undergoes regulatory review.
- Moves to marketing and post-market surveys.
However, high failure rate limits the number of drugs advancing from discovery to clinical phases.
Role of AI in Drug Development
Advantages of AI
- Time Efficiency: AI drastically cuts down the time needed for target discovery and drug-target interaction understanding.
- Increased Accuracy: Enhances the accuracy of predicting interactions between drugs and their targets.
- Cost Savings: Reduces the financial burden associated with traditional drug development methods.
AI Tools
AlphaFold and RoseTTAFold:
- Developed by DeepMind (a Google company) and the University of Washington.
- Based on deep neural networks.
- Produce three-dimensional protein structures using large amounts of input data.
Upgraded Versions:
- AlphaFold 3 (by Isomorphic Labs) and RoseTTAFold All-Atom.
- Predict structures and interactions for proteins, DNA, RNA, and small molecules.
- Use generative diffusion-based architectures.
- AlphaFold 3 has 76% accuracy in predicting interactions compared to RoseTTAFold All-Atom’s 40%.
Limitations of AI in Drug Development
Accuracy Limitations:
- Up to 80% accuracy in predicting interactions.
- Lower accuracy for protein-RNA interaction predictions.
Single Phase Contribution:
- Only aids in target discovery and drug-target interaction.
- Pre-clinical and clinical phases are still required.
Model Hallucinations:
- Occur due to insufficient training data, causing incorrect predictions.
Restricted Access:
- DeepMind has not released the code for AlphaFold 3, limiting independent verification and broad utilization.
AI in Drug Development in India
Infrastructure Needs:
- Requires large-scale computing infrastructure with fast GPUs.
- GPUs are expensive and have a quick expiration date.
Skill Shortage:
- Lack of skilled AI scientists compared to the U.S. and China.
Potential for Growth:
- India has a rich history in structural biology.
- Growing pharmaceutical industry can lead the way in AI application for drug discovery and testing.
Future Directions Of AI In Drug Discovery
- Next-Gen AI Models: Researchers are constantly developing more sophisticated AI models for drug discovery. These models could incorporate a wider range of data types (beyond just protein structures) and leverage even more powerful computing resources to achieve even greater accuracy in target identification and drug design.
- AI and Robotics Integration: The future could see a closer integration of AI with robotics in drug discovery labs. Imagine AI systems guiding robots for automated high-throughput screening of drug candidates or even designing and synthesizing new molecules based on AI predictions.
- Personalized Medicine: The ultimate goal of AI in drug discovery might be the development of personalized medicine. By analyzing a patient’s individual genetic profile and disease characteristics, AI could help design customized treatments with higher efficacy and fewer side effects.
- Focus on Undruggable Targets: Traditionally difficult-to-target molecules might become more tractable with the help of advanced AI. This could lead to breakthroughs in areas like protein-protein interactions or RNA targeting, opening doors to new classes of drugs.