Artificial Intelligence (AI) in Drug Discovery and Development

Artificial Intelligence (AI) in Drug Discovery and Development

Start

Why Drug Discovery Needs AI​

Drug discovery is:

Expensive ($1-3 billion per drug)

Slow (10-15 years)

HIgh attrition rate

Failures often occur due to:

Massive growth of chemical & biological data​

Pharmacokinetic and Pharmacological (ADMET) issues​

Poor efficacy

AI and its subset Machine Learning (ML) are now utilized in all four stages of drug discovery and development process!

AI in Drug Discovery

& Development

AI is now being used at every major stage of drug development. This includes finding disease-related targets, such as proteins, discovering drug molecules, improving how drugs are made, and running clinical trials in people.

Click on the images located in the windows to expand.

Stages:

Stage 1

Stage 2

AI in Drug Discovery & Development

Stage 3

Stage 4

Limits & Challenges

Click here

Final Message

Optional Review:​Rethinking drug design in the artificial intelligence era (est. reading time: ~25 minutes)

Optional Review:​Applications of machine learning in drug discovery and development (est. reading time: ~35 minutes)

Knowledge Check

Thank you!

Reference: C&EN (ACS) — “AI Is Taking Over Drug Discovery,” October 2025​ Slide preparation and visualization support: ChatGPT (OpenAI)

• Out of millions, roughly 700-900 proteins are targeted for FDA approved drugs​
• Drug discovery is hindered by limited structural and functional insights of new proteins.​
• Protein target structures are traditionally determined using X-ray crystallography and NMR

Stage 1: Finding Disease ​ Causing Protein Targets (Traditional)​

• AI mines literature and patient data to identify disease-associated protein targets​​

• AI predicts protein structures when experimental data are unavailable​

• AI uncovers new druggable pockets on known proteins

Stage 1: Finding Disease ​Causing Protein Targets (AI-Enabled)​

• Conventionally, medicinal chemists design, synthesize and test hundreds to thousands of molecules to identify a handful of molecules as effective protein binder​
• Traditional Drug discovery is a slow, iterative process requiring repeated optimization and testing​
• Following initial optimization, lead compounds are further refined for favorable pharmacokinetic profiling

Stage 2: Designing New Molecules​ (Traditional)​

• AI and ML tools are used to computationally simulate drug discovery experiments​
• These methods prioritize the most promising molecules from billions of possibilities​
• The strategy has the potential to shorten stage-2 from years to just weeks

Stage 2: Designing New Molecules​ (AI-Enabled)​

• Traditionally, drug development process relies on human chemists​
• Compounds are synthesized one molecule at a time​
• Process is slow and labor-intensive

Stage-3: Process Development and Manufacturing of Drug Candidates (Traditional)​

• Companies now deploy robots and AI to build automated laboratories​
• These labs conduct experiments 24/7, continuously generating data​
• AI and ML models are trained on newly generated data​
• Models identify efficient synthetic routes to target molecules​
• Enables rapid and scalable molecule production

Stage-3: Process Development and Manufacturing (AI-Enabled)​

After stage 2, drug discovery transitions into process development and manufacturing. The focus shifts from making molecules to making them efficiently, safely, and at scale through route optimization, scale-up synthesis, and manufacturing, ultimately yielding a final drug product suitable for clinical use.​

Stage 3: Process Development and Manufacturing of Drug Candidates​

Example

• AI/ML models streamline patient identification and enrollment​
• AI assists in clinical trial protocol design​
• Digital tools help educate, engage, and retain participants​
• Enables faster data analysis and more efficient trials

Stage 4: Design & Execution of Human Clinical Trials (AI-Enabled)​

The final stage is testing drugs in humans. Traditionally, researchers design protocols, recruit participants, and analyze trial data manually. This process is slow, expensive, and often a major cause of delays in drug development.

Stage 4: Design & Execution of Human Clinical Trials (Traditional)​

Limits & Challenges

Limited clinical impact: Few genuinely AI-designed drugs have reached successful clinical outcomes​ Lack of true novelty: Many AI-identified targets were already known or therapeutically explored​ Data limitations: AI performance is constrained by insufficient, biased, or low-quality biological data​ Translation gap: Models may produce plausible predictions that fail to translate to clinical efficacy

Take Home Message

AI is a powerful tool, not a magic solution​. Best results come from:​

• AI + human knowledge (chemistry & biology), experience and intuition​
• Understanding AI concepts is essential for modern day drug discovery and development

Try again!

AI helps analyze massive biological and chemical datasets to improve decision-making and accelerate the drug development process.

Incorrect

AI increases speed and efficiency but still requires human oversight and expertise.