‘AI lowers the capital needed’: India’s drug discovery push faces test of human trials

AI acts as a filtering and design tool in early drug discovery. It can examine biological data, identify promising targets, screen many possible compounds and suggest molecular designs. This matters because researchers can focus limited time and money on the candidates most likely to work.

At Abhinavayan Biotech, AI helps smaller companies select targets, screen compounds and design molecules before laboratory experiments. Peptris also uses AI to run more than a dozen potential drug programmes with a team of about a dozen people. The basic mechanism is prediction: software compares patterns in data and ranks molecules for further testing.

AI does not make drug discovery easy or remove the need for proof. Santanu Datta of Bugworks said AI works best with large amounts of reliable data. The technology can reduce the number of compounds researchers make and test, but laboratory studies, preclinical work and human trials still determine whether a candidate is safe and effective.

The clearest time comparison is Peptris’s experience. The company moved one programme to a late preclinical stage in about 18 months, compared with more than four years using conventional methods. That suggests AI can substantially shorten early discovery, although one company’s result is not a universal industry benchmark.

The article does not give a precise rupee amount or percentage for savings. It explains the mechanism instead. Cheaper computing and AI tools can reduce the number of compounds researchers need to make and physically test. Smaller teams can therefore investigate more candidates while spending less on early laboratory work.

The savings have limits. Vishesh Rajaram said AI lowers the capital needed to create a drug asset, not the capital needed to prove it works in patients. Human trials, clinical infrastructure and long-term development remain expensive. AI changes the economics of early discovery, but it does not make the whole process cheap or predictable.

Late preclinical development is still before human testing. At this stage, researchers have gathered evidence from laboratory studies and other non-human work, but they have not established how the candidate behaves in people. The result is encouraging evidence, not a patient-ready treatment.

A drug must still pass carefully controlled clinical trials. Researchers need to test safety, dosing, side effects and possible benefits in humans. A compound can look promising in laboratory models yet fail because it is toxic, does not reach the right tissue, is poorly absorbed or does not improve patients’ health.

That gap is central to the article’s warning. Moderna co-founder Noubar Afeyan said India needs preclinical and clinical capabilities established at scale, rather than relying on AI alone. Peptris’s rapid progress shows what AI can accelerate. It does not show that the candidate will succeed in human trials or receive approval.

India needs a connected system that can move candidates safely from discovery into human testing. That includes well-equipped preclinical laboratories, facilities able to produce consistent clinical-trial material, hospitals and trial sites, trained investigators, ethics review, regulators, patient recruitment and systems for collecting reliable data. These requirements matter because AI can identify candidates, but people and institutions must test them.

The article names the broad gap rather than listing every facility. Noubar Afeyan said India needs much more infrastructure before and during human trials. He argued that preclinical and clinical capabilities at scale are probably more important than AI alone. In practice, that means enough laboratories and trial capacity to handle many programmes consistently.

India already has sophisticated scientists and contract research organisations, according to Peptris’s co-founder. The challenge is extending that strength into company-owned medicines and human development. More capacity could let promising candidates advance faster, while strong oversight and data systems would help separate genuinely useful drugs from attractive but unsuccessful predictions.

The article does not state how many experimental drugs entering Phase 1 eventually receive approval. A commonly cited industry estimate is roughly 10 to 15 percent, although the rate varies by disease area, drug type and how success is measured. This figure comes from established industry analysis, not from the article.

Phase 1 mainly tests an experimental drug’s safety, tolerability and dosing in people. Later studies must provide stronger evidence of benefit, compare outcomes in larger groups and satisfy regulators. Candidates can fail at any step, even after promising laboratory and early clinical results. Approval is therefore the exception, not the expected outcome for each programme.

For investors, this makes drug discovery a portfolio and patience game. AI may reduce early costs and help teams choose candidates more intelligently. It cannot remove biological uncertainty or the expense of clinical proof. Companies need patient capital, sustained commitment and a long-term vision, as Megha Goyal told CNA.

India’s pharmaceutical sector historically became strongest in generic medicines, manufacturing and contract research for international firms. Those areas built major industrial and scientific capabilities without requiring every Indian company to finance an original medicine from discovery through approval. The article contrasts this record with India’s less-developed position in original drug discovery.

Original medicines require long development timelines, expensive experiments and a willingness to absorb failure. Peptris’s Venkatasubramanian Narayanan said Indian scientists and contract research organisations already perform sophisticated discovery work. However, Indian companies have traditionally taken on less of the financial risk involved in owning the medicines being developed.

That pattern may now be changing. Venture investor Vishesh Rajaram said more Indian scientists are founding companies around their own novel chemistry, biologics and AI-first programmes. Government funding is also growing, including about 16 billion rupees for 41 pharmaceutical and medical technology research projects. The shift still depends on patient capital and clinical infrastructure.

A clinical trial is a planned study that gives an experimental treatment to human volunteers or patients under defined conditions. Researchers monitor safety, side effects, dosing and signs of benefit. Later-stage trials usually involve more participants and provide stronger evidence about whether a treatment helps patients. This human evidence is needed before regulators can consider approval.

A drug can behave differently in people because laboratory models simplify biology. The human body absorbs, distributes, breaks down and removes substances in complex ways. Differences in metabolism, immune responses, genetics, disease and interactions with other medicines can change both benefit and toxicity. A molecule that affects a target in a dish may therefore fail to produce the same effect in patients.

The article’s warning reflects this gap. AI can identify promising compounds and accelerate preclinical work, but it cannot prove that an asset works in patients. That is why Noubar Afeyan emphasised building preclinical and clinical capabilities at scale, alongside discovery tools.