Artificial intelligence is beginning to compress the early stages of cancer drug discovery from nearly eight years to just three or four months, according to oncologist and Manas AI co-founder Dr. Siddhartha Mukherjee. Speaking at the Hindustan Times Leadership Summit 2026, he described how AI-powered molecular design combined with laboratory automation can accelerate the search for potential cancer treatments. The development could change how pharmaceutical researchers identify and develop experimental drug candidates.
However, the breakthrough concerns early-stage drug discovery rather than the entire process of developing and approving a medicine. Clinical trials, safety testing, manufacturing and regulatory reviews still take considerable time. Mukherjee said AI could help researchers reach promising drug candidates much faster, but the technology has not eliminated the challenges involved in proving that a medicine is safe and effective in humans.
How AI Is Accelerating Cancer Drug Discovery
Mukherjee, an oncologist and co-founder and CEO of Manas AI, explained that the company’s approach combines machine learning with the established principles of physics and chemistry.
Traditional drug discovery often involves identifying a disease-related biological target and searching for molecules capable of interacting with it. Researchers may need to examine enormous numbers of chemical compounds before finding a candidate with the desired properties.
Manas AI is attempting to change this process by using algorithms to identify vulnerable regions on disease-causing proteins and design molecules that can interact with those regions.
Mukherjee compared the process to finding a keyhole in a lock. Once researchers identify the right location on a protein, AI can help design a molecule that fits that biological target.
Instead of relying entirely on the screening of existing compounds, the system can generate new molecular structures using individual atoms and chemical principles.
This approach is intended to reduce the time required to move from identifying a biological target to producing an experimental drug candidate.
From Years of Research to Three or Four Months
Mukherjee highlighted the difference between traditional research timelines and the company’s AI-assisted approach.
He said that work which previously required approximately seven to eight years could now be completed in three to four months in the early discovery process.
One example involved daraxonrasib, an experimental pancreatic cancer drug. Mukherjee said identifying a particular target region on a pancreatic cancer-related protein had taken human researchers approximately 10 years.
According to him, Manas AI’s algorithm identified the same target region in just two hours using its specialised target-identification system.
The comparison illustrates how computational methods can accelerate specific research tasks. It does not mean that the complete development of daraxonrasib, or any other cancer medicine, took only two hours.
Mukherjee also described how Manas AI sends batches of approximately 50 to 100 compounds to chemical companies in China for laboratory testing. The companies return experimental data in roughly three weeks, allowing researchers to assess the molecules and refine subsequent designs.
This combination of computational design and physical laboratory testing creates a faster development cycle than relying on computational predictions alone.
AI-Designed Molecules Show Promising Early Results
Mukherjee also discussed a molecule designed through the company’s system without direct human intervention during the initial molecular-generation process.
He claimed that the experimental molecule was approximately 200 times more potent than a competing medicine available in the market.
This is a company-reported experimental comparison, not evidence that the candidate is 200 times more effective in patients. Laboratory potency measurements do not automatically translate into better clinical outcomes.
A potential cancer medicine must demonstrate that it can reach the intended biological target, produce the desired effect and avoid unacceptable toxicity.
Researchers must also establish how the drug behaves inside the human body, including how it is absorbed, distributed and eliminated.
Nevertheless, the ability to generate and test more experimental candidates in shorter periods could help pharmaceutical researchers explore a wider range of possible treatments.
Why AI Cannot Replace Clinical Trials
Despite the progress in early drug discovery, Mukherjee cautioned that AI has not solved the biggest challenges in bringing medicines to patients.
Once a promising molecule is identified, it must undergo preclinical testing and human clinical trials before regulators can consider approving it.
Clinical development generally involves several stages:
| Development stage | Main objective |
|---|---|
| Early discovery | Identify biological targets and design potential drug molecules |
| Preclinical testing | Evaluate biological activity, toxicity and safety before human testing |
| Phase I trials | Assess initial safety, tolerability and dosage |
| Phase II trials | Examine effectiveness and continue safety evaluation |
| Phase III trials | Compare treatment outcomes in larger patient populations |
| Regulatory review | Evaluate evidence before potential approval |
AI can assist researchers in several of these stages, including analysing biological data, designing molecules and helping plan clinical studies.
However, it cannot simply replace the need to test medicines in people.
Mukherjee highlighted the uncertainty associated with Phase II clinical trials, where promising experimental treatments can still fail to demonstrate sufficient effectiveness.
A molecule that performs exceptionally well in a laboratory may not produce the same results in patients. Human biology, disease variation, drug resistance and unexpected side effects can all affect outcomes.
This remains particularly important in cancer research, where different cancers—and even different patients with the same cancer type—can respond differently to treatment.
Why India Could Benefit From AI-Powered Drug Development
Mukherjee also argued that India has an opportunity to become an important participant in the next phase of pharmaceutical innovation.
The country has a substantial pharmaceutical manufacturing industry, a large healthcare research ecosystem and experience producing medicines at relatively competitive costs.
AI could strengthen these advantages by reducing the time and resources required to identify potential drug candidates.
However, the next challenge is making the entire development process more efficient. Faster molecular design will have limited impact on the final cost of medicines if clinical trials, manufacturing and regulatory procedures remain expensive and time-consuming.
Mukherjee suggested that India and China could benefit from reforms that make drug testing and development more efficient without weakening patient protections.
The objective would be to reduce unnecessary delays while preserving the scientific standards required to establish whether a medicine works.
The Bigger Picture
The emergence of AI-powered drug discovery represents a potential shift in pharmaceutical research. Instead of depending exclusively on lengthy cycles of conventional laboratory screening, researchers can increasingly use computational systems to identify biological targets, generate new molecules and prioritise candidates for testing.
The technology could allow research teams to examine more possibilities with fewer resources and accelerate the early stages of drug development. This may be particularly valuable in cancer research, where the complexity of biological targets makes discovering effective treatments difficult.
However, faster discovery should not be confused with faster medical approval. The most important test is whether AI-designed molecules can consistently demonstrate safety and meaningful benefits in clinical trials. The technology’s long-term contribution will depend on its ability to produce medicines that improve patient outcomes, not simply generate promising molecules more quickly.
Looking Ahead
Manas AI’s approach highlights how combining machine learning, molecular design and automated laboratory testing could reshape the early stages of pharmaceutical research. The next major milestone will be demonstrating that candidates developed through these systems can successfully progress through human trials. Independent validation, published research and clinical results will be important in assessing the actual effectiveness of the approach.
For patients, the potential benefit is significant: more efficient research could eventually help bring new treatments within reach, including for cancers that remain difficult to treat. But the journey from an AI-generated molecule to an approved medicine remains long, and clinical evidence will ultimately determine whether these technological advances translate into better cancer care.
Get the day’s top stories in your inbox
One concise email. No spam, unsubscribe anytime.
