Alex Zhavoronkov, PhD, is the founder and CEO of Insilico Medicine, a leading clinical-stage biotechnology company developing next-generation generative artificial intelligence and automated platforms for drug discovery. Since 2014, he has invented critical technologies in the field of generative artificial intelligence and reinforcement learning (RL) for the generation of novel molecular structures with the desired properties and the generation of synthetic biological and patient data. Under his leadership, Insilico raised over $500 million in multiple rounds from expert biotechnology, healthcare, and financial investors, opened R&D centers in 6 countries and regions, and partnered with multiple pharmaceutical, biotechnology, and academic institutions. Since 2021, the company nominated more than 20 preclinical candidates, started 6 human clinical trials, and entered Phase II with an AI-discovered novel target and AI-designed novel molecule. Since 2012, he has published over 200 peer-reviewed research papers and 3 books. He serves on the advisory or editorial boards of Trends in Molecular Medicine, Aging Research Reviews, Aging, and Frontiers in Genetics, and founded and co-chairs the Annual Aging Research and Drug Discovery (11th Annual in 2024). He is the adjunct professor of artificial intelligence at the Buck Institute for Research on Aging.

Ten years ago this was science fiction, and now we have AI-discovered targets and AI-designed molecules progressing through clinical trials.

You’ve popularized the term Pharmaceutical Superintelligence (PSI). What do you mean by it?

The topics of human-level artificial general intelligence (AGI) and artificial superintelligence (ASI) have captivated researchers for decades, dramatically more so in recent years, with much disagreement on when and if it is possible. “Solve AI and it will solve everything else” is a popular assumption and goal for these endeavors.

Applying this vision to the healthcare and pharmaceutical industries, we can define the pharmaceutical superintelligence (PSI) as a fully-autonomous platform capable of discovering and designing a perfect small molecule or a biologic drug, together with the biomarker for patient selection, producing a significant disease-modifying or curative response for any disease without failure and without the need for further human experimentation.

Where did the idea come from? How is Insilico Medicine evolving into building Pharmaceutical Superintelligence?

When I started Insilico in 2014, we were working on biomarkers and aging clocks, essentially ways to measure biological age and disease risk. As AI has advanced, especially generative models, we realized we could pivot into drug discovery itself. Over time, we built out modules that solved specific problems, like PandaOmics for target discovery and Chemistry42 for molecule design, and then we started linking them together. The vision of Pharmaceutical Superintelligence came from the idea of what if these pieces were fully integrated, validated, and able to learn from each other continuously? That’s how you will transform the entire industry.

Insilico has made headlines for compressing development timelines. Can you share some milestones that demonstrate this?

Traditionally, drug development takes about 10+ years from target to market, with preclinical development alone taking approximately 4.5 years. With AI and experimental feedback, we’ve shown the preclinical development can be done much faster. For example, our QPCTL program went from discovery to DC stage in just nine months and our TNIK candidate to DC stage in 18 months. Today, we have 40+ programs, including one for idiopathic pulmonary fibrosis that showed positive Phase 2a results. Those milestones prove that AI isn’t just theoretical but it’s producing tangible therapeutics.

If validation is the bottleneck, what unlocks the next phase of PSI?

Four levers. First, open program-level benchmark repositories that tie omics, chemistry, and clinical outcomes together. Second, distilling validated single-task “teacher” models into versatile multimodal agents. Third, pan-flute simulation cascades, cheap filters first and physics-heavy models last. Fourth, community reinforcement learning from experimentally verified feedback, so the whole ecosystem learns from real programs.

Can you unpack that “teacher–student” idea?

Pharma models often “expire” while we wait for validation. The fix is continuity as validated task models become teachers that generate trustworthy synthetic data and provide guardrails while training more capable, multimodal students. You can preserve hard-won intelligence while upgrading capabilities.

Let’s talk about IPF. Why was that such an important moment for Insilico?

IPF is a devastating lung disease, often fatal within a few years, and patients have very limited options. Our program started entirely with AI, PandaOmics identified TNIK as a novel target, Chemistry42 designed the molecule, and our team advanced it into the clinic. The Phase 2a trial showed a +98 mL improvement in lung function over placebo and the results were published in Nature Medicine. It wasn’t just a success for us but was the first real proof in the world that our AI-driven molecules can show efficacy in patients.

How does aging biology fit into program, beyond single diseases?

In Insilico’s IPF program, AI’s contribution to biology extended far beyond simply identifying a novel target for IPF. Crucially, it also uncovered numerous potential biomarkers linked to aging processes, which are vital for advancing translational research and providing critical reference points for further clinical trials.

You’ve compared pharma to autonomous driving and even world models. Why?

Validation at scale matters and autonomous driving accumulates millions of rides. Pharma needs analogous “trips” across programs and trials. As we collect them, we can build better world simulations for drugs, so agents learn from richly simulated clinical reality while we continue to validate in the real world.

How realistic is your prediction that the first fully AI-designed drugs will be available to patients in the next five to six years?

I would be surprised if we didn’t see it in that timeframe. We already have more than 40 internal programs, with the most advanced program having completed Phase IIa with promising safety and efficacy validation. Future success depends on continued trial outcomes, regulatory review, and safety. I hope Insilico will be the first, but even if it’s not us, someone will cross that line before 2030. The important thing is that the field has reached a point where that’s a realistic expectation.

What should scientists and companies do tomorrow to get closer to PSI?

Like with other AI systems, both training and trust come from validation. Many models that have been validated for specific tasks in simulation, preclinical, and clinical experiments can be used to train larger, more capable, multimodal LLMs. The validated models can generate synthetic data for supervised training of the next generation of LLMs, with some computationally efficient models being used directly in the reinforcement learning loop.

There are several pathways for achieving PSI. The most likely path is the convergence of the highly-capable LLM models trained on Internet-scale data with significant reasoning capabilities and the highly-validated, smaller, task-specific, multi-modal biology-, chemistry-, and physics-based models based on either advanced and relatively primitive architectures. This convergence might also happen with LLMs that use such specialized models and tools in the agentic workflows, akin to medicinal chemists using traditional drug design software and tools.

The key to PSI is the ability to use the validated “legacy” models for synthetic data creation, or use for reinforcement learning and benchmarking the next-generation models similar to how knowledge and intelligence is passed from generation to generation in human society.

Looking ahead, what gives you the most optimism about Pharmaceutical Superintelligence?

We’re already seeing the early signs of PSI in action. Ten years ago this was science fiction, and now we have AI-discovered targets and AI-designed molecules progressing through clinical trials. We’ve compressed preclinical timelines from the traditional 4.5 years to under 18 months in some cases and built benchmarks across more than 22 programs. Every program’s real-world data helps us train better models. That loop of evidence gives me confidence we’ll get to a point where AI is reliably driving drug discovery end to end.

The post Q&A with Alex Zhavoronkov, CEO of Insilico Medicine appeared first on citybiz.

This contant was orignally distributed by citybiz. Blockchain Registration, Verification & Enhancement provided by NewsRamp™. The source URL for this press release is Q&A with Alex Zhavoronkov, CEO of Insilico Medicine.