Insilico Medicine Links AI Platforms to Speed Up Drug Discovery

In a development that could reshape how new medicines are found, Hong Kong and New York-based Insilico Medicine has demonstrated a way to link its artificial intelligence platforms for target discovery and generative chemistry. The result: a novel drug target and a brand-new molecule designed to hit it, all in a fraction of the time traditional methods would take.

The company published its findings in the journal Nature Biotechnology, outlining how it stitched together two of its proprietary AI systems. The goal was to prove that the entire early-stage drug discovery pipeline — from identifying what to target to designing a molecule that can act on it — can be run through a single, coherent AI workflow.

Two Platforms, One Workflow

Insilico's first platform, PandaOmics, is built to sift through vast biological datasets to uncover novel targets — the proteins or genes that, if modulated, could treat a disease. The second, Chemistry42, is a generative chemistry engine that can dream up entirely new molecular structures that bind to those targets.

Until now, these systems operated largely in isolation. Researchers would use one to find a target, then manually hand off that information to chemists who would design molecules. Insilico's latest work shows that this handoff can be automated, with the AI platforms talking to each other directly.

Proof in a Fibrosis Target

The team focused on idiopathic pulmonary fibrosis, a chronic lung disease with few treatment options. PandaOmics scanned gene expression data and other biological signals to pinpoint a target that had not been widely pursued in drug development. The AI flagged a specific protein as a promising candidate.

That target was fed directly into Chemistry42. The generative engine produced a list of novel small molecules predicted to bind to the target with high affinity. The entire process, from target discovery to molecule design, took 46 days. Traditional drug discovery, even at its fastest, can take months or years to reach this stage.

Validated in Lab Tests

Insilico did not stop at the computer. The company synthesized the top-ranked molecule and tested it in cellular and animal models of fibrosis. The results showed that the molecule engaged the intended target and reduced fibrosis markers.

Alex Zhavoronkov, CEO of Insilico Medicine, called the work a step towards 'end-to-end AI-driven drug discovery.' He said the approach could cut the time and cost of finding new drug candidates, especially for diseases where targets are poorly understood.

The study has limitations. The molecule has not yet been tested in human trials, and whether it will prove safe and effective in people remains unknown. But the speed and novelty of the target-molecule pair are what caught the attention of researchers in the field.

What This Means for Drug Development

The pharmaceutical industry has been experimenting with AI for years, but most efforts have focused on using AI for one step — say, predicting protein structures or screening existing compounds. Insilico's work is among the first to show a seamless, automated link between target discovery and molecule generation.

Other companies, including Recursion Pharmaceuticals and Exscientia, have also used AI for drug discovery, but each uses a different approach. Insilico's claim of a fully connected pipeline, from target to molecule in under two months, sets a new benchmark for speed.

Critics point out that the true test will come in clinical trials, where most drug candidates fail. But the ability to rapidly generate novel hypotheses and test them in the lab could accelerate the early stages of research, especially for rare or neglected diseases where traditional drug development is often too slow or too expensive.

Insilico's next steps will involve moving its fibrosis molecule into more advanced preclinical testing and expanding the linked platform to other diseases. The company is also exploring partnerships with larger pharmaceutical firms interested in using the AI pipeline for their own targets.