How self-driving labs are changing drug development

Automation is changing the face of scientific research. In drug discovery, where timelines can stretch over a decade and the cost of bringing a single therapeutic to market can exceed a billion pounds, the autonomous laboratory offers a practical route to greater efficiency. Robotic systems capable of running experiments around the clock, generating data in real time and feeding results back into machine learning models are changing how discovery is carried out in the modern laboratory.

At the forefront of this movement is a new generation of self-driving laboratories that merge automated synthesis, biological screening and computational intelligence into a seamless cycle of discovery. These systems are designed not merely to execute experiments, but to learn from them – using data-driven feedback loops to refine their approach and accelerate progress from hit identification to lead optimisation.

One of the scientists helping to bring this vision to life is Stuart R Green, a Staff Scientist at the Acceleration Consortium (AC) at the University of Toronto. Within the consortium’s Medicinal Chemistry Self-Driving Lab (SDL), Stuart plays a pivotal role in developing automated biophysical assays and closed-loop workflows that connect chemical synthesis with biological evaluation. His work is focused on enabling a faster, more intelligent approach to small molecule discovery.

“The chemical probes we work on are important starting points in drug development,” he says. “They act as highly selective, potent small molecule binders of proteins of interest.”

Breaking the hit-to-lead bottleneck

In modern drug discovery, the earliest stages of research are often the most unpredictable. While advances in high-throughput screening have enabled scientists to identify potential hits from vast chemical libraries, the subsequent process of refining those hits into viable leads remains time-consuming and expensive.

Currently in the drug discovery pipeline there is a major bottleneck at the hit-to-lead optimisation stage. Initial hit candidates can be identified increasingly rapidly, but these hits are typically weak and may have poor selectivity.

Stuart describes the challenge: “Currently in the drug discovery pipeline there is a major bottleneck at the hit-to-lead optimisation stage. Initial hit candidates can be identified increasingly rapidly, but these hits are typically weak and may have poor selectivity.”

Traditionally, medicinal chemists would synthesise and test hundreds of analogues to establish the structure–activity relationship of a compound. This iterative cycle can take months or even years. The SDL’s approach seeks to streamline the process by integrating synthesis, screening and data analysis into one continuous automated system.

“Our approach aims to bypass these restrictions by constraining the search space to compounds that can be synthesised from a set of diverse building blocks in a robust set of reactions,” Stuart explains. “We perform AS-MS assays without compound purification in a direct-to-biology workflow on a fully autonomous system working in a closed loop.”

This direct-to-biology methodology means that up to 100 compounds can be synthesised and characterised simultaneously, drastically reducing the time between idea and result. Crucially, the system is designed to learn from its own output. Experimental data are used to train machine learning models that inform the next cycle of synthesis, gradually refining compound design until predefined potency and selectivity targets are achieved.

Building an autonomous ecosystem

Constructing a self-driving lab from scratch requires more than simply purchasing robotic equipment. It involves creating an integrated ecosystem in which each component – from synthesis robots to analytical instruments – operates seamlessly.

“Obtaining a chemistry-capable liquid handler able to perform chemical synthesis in an inert atmosphere free from humidity with a variety of organic solvents outside of a glove box was challenging,” Stuart recalls. “Meeting these performance demands and addressing safety requirements for ventilation meant that early on we realised a dedicated liquid handler for carrying out chemical synthesis would be needed, that was separate from a secondary liquid handler, for dispensing the aqueous solutions needed for biochemical assay preparation.”

After extensive consultation with instrument manufacturers, the team developed a dual-handler setup that met both performance and safety requirements. “Finding the optimal solution to meet our needs required extensive research and discussions with vendors in advance of instrument purchase,” Stuart says. “In the end, our discussions with vendors allowed us to develop a customised solution that met our immediate needs with room for future expansion.”

Discovery through automation

Automation has long been used in biological screening, but its application to chemical synthesis and assay integration is comparatively recent. For Stuart, the benefits extend beyond productivity. Automation enhances reproducibility, standardisation and data quality – key issues in early-stage discovery where subtle variations can skew results.

An automated lab setup is particularly valuable in the screening of small molecule libraries due to the highly repetitive nature of the task.

“An automated lab setup is particularly valuable in the screening of small molecule libraries due to the highly repetitive nature of the task,” he explains. “While the development of wholly in silico tools for de novo hit prediction remains a lofty long-term goal, at present these approaches have met with limited success and there is no consensus on the best approach.”

By using automated systems to screen inhibitors against panels of related proteins, the SDL is able to tackle one of medicinal chemistry’s most persistent problems: selectivity. “Working in parallel with multiple related proteins simultaneously would be challenging in a traditional lab owing to the large amount of manual pipetting work and interpreting the large amount of data generated,” he says. “Looking at multiple protein family members at once also allows for early identification of compounds with poor selectivity through automated data analysis modules.”

Intelligence meets experimentation

As artificial intelligence becomes more deeply integrated into laboratory workflows, the boundaries between computation and experimentation are blurring. Stuart believes that AI will eventually change how chemists think about molecular design altogether.

Since AI agents can infer patterns in a higher dimensional space from massive datasets better than a human chemist can, it can potentially suggest molecules that would not be immediately obvious to human intuition.

“Since AI agents can infer patterns in a higher dimensional space from massive datasets better than a human chemist can, it can potentially suggest molecules that would not be immediately obvious to human intuition,” he says. “This means that not only could AI increase the speed at which candidate compounds are proposed but it could also make suggestions that would otherwise be overlooked.”

The long-term goal is a fully autonomous discovery engine in which experimental data continually refine predictive models, creating an evolving partnership between human insight and machine reasoning.

Lessons for labs new to automation

For research groups exploring automation for the first time, the task can appear overwhelming. Yet, as Stuart points out, the ecosystem of commercial tools supporting automation has expanded rapidly in recent years.

“In applications involving the use of standardised labware, commercial tools have already been developed to facilitate automation of most of these processes, ranging from liquid handling, plate sealing and movement, barcoding, centrifugation and plate reading,” he explains.

“Before beginning my current position, I was unaware of how many tools had already been developed in this space and would recommend to anyone interested in laboratory automation that they discuss with individuals who already work with an automated lab setup.”

When investing in new equipment, he advises teams to think beyond the instruments themselves and consider how they will be orchestrated. “When purchasing instruments, it is important not just to understand their physical capabilities, but also how they will be operated autonomously,” he says.

Many labs opt for commercial orchestration platforms, but others choose to develop bespoke solutions. “This can certainly be more challenging and may require dedicated personnel for software development but it can allow more advanced customisation of workflows and fine-grained control over all the processes involved,” Stuart notes.

The future of accelerated discovery

Looking ahead, the impact of automation on early drug discovery could be profound. By reducing both the time and cost of hit-to-lead optimisation, self-driving labs have the potential to broaden the scope of research, enabling scientists to pursue targets that were previously overlooked for economic reasons.

Delegating both the manual labour associated with running experiments to an automated lab setup and the mental labour of compound selection in a closed loop automated workflow will help to reduce this barrier.

“Time and cost constraints are a major barrier to the development of novel drugs,” Stuart says. “Delegating both the manual labour associated with running experiments to an automated lab setup and the mental labour of compound selection in a closed loop automated workflow will help to reduce this barrier.”

These efficiencies could facilitate progress towards treatments for rare diseases and conditions prevalent in low-income regions, areas that have historically received limited research attention.

“This will allow drug candidates to be developed for rare diseases that were previously not considered due to economic reasons, or potentially find treatments for diseases mainly associated with the developing world,” he says. “Reducing time and cost associated with chemical probe discovery will also enhance our knowledge of the biology of poorly characterised proteins by accelerating the production of high-quality chemical tools to interrogate their function.”

A new model for scientific discovery

As automation, machine learning and chemical biology come together, the self-driving lab is changing how research is organised and carried out. It does not replace human expertise but extends it, allowing scientists to work more efficiently and test ideas at a greater scale.

In this area of research, Stuart R Green’s work shows how automation is changing early drug discovery. The self-driving lab is no longer theoretical – it is already altering how new medicines are found.

As automated workflows expand across drug discovery, integrating reliable imaging into robotic environments is becoming an essential part of building scalable laboratory systems. Our upcoming webinar explores these challenges in detail, presenting a real-world case study of integrating high-content imaging into a fully automated robotic WorkCell.

Meet the expert

Stuart R Green is a Staff Scientist at the Acceleration Consortium (AC) working in the Medicinal Chemistry Self-Driving Lab research group. He plays a critical role in the development of biophysical assays for automated drug discovery applications and hit validation.

Before joining the AC, he completed his post-doctoral research at the Structural Genomics Consortium at the University of Toronto under the supervision of Masoud Vedadi and Levon Halabelian. His work there focused on characterising protein–small molecule interactions through the design of high-throughput screening assay campaigns for hit discovery. Central to these projects was the optimisation and implementation of a wide range of biophysical assays including fluorescence polarisation (FP)-based displacement assays, surface plasmon resonance (SPR) assays, differential scanning fluorimetry (DSF) thermal shift assays and others.

Stuart completed his PhD at Carleton University in the laboratory of Ken Storey, where he researched the regulation of enzymes governing central metabolic pathways in animal models of hypometabolism.