Researchers develop a light-activated drug discovery platform for next-generation cancer and antimicrobial treatments using automation

Imperial-led researchers used an automated approach to create a library of metal-based compounds that have the potential to accelerate the discovery of cancer and antimicrobial therapies.

Finding new medicines is often a slow and costly process, requiring researchers to test large numbers of compounds before identifying promising candidates.

Now, researchers, led by Professor Ramon Vilar from the Department of Chemistry at Imperial College London, and in collaboration with Professor Kenneth Kam-Wing Lo at City University of Hong Kong, Professor Nils Metzler-Nolte at Ruhr University Bochum and Professor Heather J. Kulik at the Massachusetts Institute of Technology, have shown how chemistry and automation can work together to speed up this search. By studying hundreds of metal-based compounds and training machine learning models on the results, the team created a platform to discover future treatments for diseases, including cancer and drug-resistant infections.

"Metal-based drugs have huge potential, but the field has traditionally explored them a handful of compounds at a time. Automation lets us make and test hundreds of compounds in parallel, and machine learning lets us see the patterns in that data, telling us where to look next." – Dr Tim Kench

Exploring the potential of metal-based medicines

Metal-based compounds have long been used in medicine, but they remain an underexplored area of drug discovery. Whilst conventional drugs are mostly built as flat two-dimensional scaffolds, metal complexes can hold several building blocks in a distinct 3D manner. Swapping out these building blocks gives researchers a modular way to fine tune a compound’s shape and properties, and so control how they behave in biological systems.

To explore that potential, the team built a library of 336 iridium-based compounds designed for photodynamic therapy, a treatment approach in which medicines are activated only when exposed to a specific wavelength of light. Once activated, these compounds generate highly reactive forms of oxygen that can damage nearby cancer cells. Because the effect is localised to the illuminated area, photodynamic therapy offers a more targeted approach than treatments that act throughout the body. This approach is also promising for fighting infections, as this light-activated damage is much less likely to cause bacteria to develop drug resistance, which is a major future health concern.

The researchers tested the compounds against cancer cells, healthy cells and bacteria, and used automated experiments to understand how they behaved in biological systems, including how readily they entered cells, where they travelled within them, and how strongly they responded to light.

They found that several compounds became more potent against cancer cells when activated by light. They also found compounds capable of triggering biological signals associated with immunogenic cell death, a process that can help alert the immune system to the presence of tumours.

The analysis also revealed an important insight: the chemistry associated with effective cancer treatments was often different from the chemistry associated with effective antimicrobial agents. This allowed the researchers to map distinct regions of chemical space linked to different therapeutic applications and develop design rules for future medicines.

"What we've built here is a platform, not a one-off result. The same approach could be pointed at a different metal, a different disease, or a different property we want to optimise." – Professor Ramon Vilar

A platform for future drug discovery

But the study was about more than discovering potential new therapeutics. By systematically testing hundreds of molecules under the same conditions, the researchers created a high-quality dataset that could be used to train machine-learning models. Those models were then used to screen more than 200,000 virtual compounds, helping the team identify the most promising candidates to synthesise and test in the lab.

Although the compounds identified are still at an early stage of development, the researchers believe that the platform could transform the way new medicines are discovered. By combining automation, biological screening and machine learning, researchers can explore vast regions of chemical space far more efficiently than traditional approaches. Rather than synthesising and testing compounds one by one, researchers can rapidly identify and prioritise the most promising drug candidates for further development.

Read the full paper titled Exploring chemical space for iridium(iii) complexes: a direct-to-biology (D2B) approach to identifying anticancer and antibacterial agents