Artificial intelligence and automation are speeding up new drug discovery
Drug discovery has always been a long, expensive, and uncertainty-filled process. Developing a single drug can take over a decade and cost billions of dollars, with absolutely no guarantee the result will ever reach the patient. But that landscape is changing, and pretty fast.
At the heart of this transformation is the DMTA cycle, which stands for Design-Make-Test-Analyze — essentially designing, synthesizing, testing, and analyzing compounds in the search for drug candidates. What used to be a linear and slow process is becoming something far more dynamic, connected, and data-driven.
And that is exactly where artificial intelligence and automation come into play — not as some futuristic trend, but as real tools that are already inside the lab. 🔬
To understand how all of this works in practice, two experts from TTP, a British company with decades of experience in lab automation and technology development for the pharmaceutical industry, shared their views on the present and future of this space.
Jon Wingfield, senior business development lead and former president of SLAS (Society for Laboratory Automation and Screening), brings nearly 30 years of experience in pharmaceutical research and development. He works closely with pharmaceutical, biotech, and life sciences organizations to accelerate innovation across the drug discovery pipeline. His expertise spans lab automation, high-throughput screening, and DMTA workflows.
Wenshu Xu, who leads the Drug Discovery Tools team at TTP, drives the development of technologies designed to speed up drug discovery through advanced automation, high-throughput experimentation, and autonomous lab systems. She has a particular interest in the concept of physical AI, where computational predictions are rapidly validated through automated experimentation in the lab.
What they have to say goes well beyond the AI hype. 👇
What is the DMTA cycle and why does it matter so much
Before diving into artificial intelligence and automation, it helps to understand what the DMTA cycle actually is and the role it plays within the pharmaceutical industry. Think of it this way: every time a scientist wants to test whether a molecule could become a drug, they need to go through four fundamental steps.
- First, they design the compound based on scientific hypotheses.
- Then, they physically make (synthesize) that molecule in the lab.
- Next, they run tests to evaluate its efficacy and safety.
- Finally, they analyze the collected data to decide the next step.
This cycle repeats hundreds, sometimes thousands of times, until a promising candidate appears on the horizon.
The historical problem is that each of these steps used to be done in isolation, with different teams, different systems, and very little real-time exchange of information. The result was a fragmented process where data generated in one step took weeks to influence the next decision.
According to Jon Wingfield, this gap between generating and using data has always been one of the biggest bottlenecks in drug discovery. Entire projects lost months because of a misinterpreted result or an analysis that arrived too late to change the direction of experiments.
Receive the best innovation content in your email.
All the news, tips, trends, and resources you're looking for, delivered to your inbox.
By subscribing to the newsletter, you agree to receive communications from Método Viral. We are committed to always protecting and respecting your privacy.
An important shift in perspective
A revealing detail shared by Wingfield is that historically, the DMTA cycle was seen as something that started with the design phase. Today, however, there is growing recognition that analysis may actually be the starting point. Advances in AI-enabled target identification, protein structure prediction, molecular docking, and binding analysis provide valuable insights before molecules are even designed.
Another important point is the value of failed experiments. Many AI models rely heavily on published literature, but unsuccessful experiments are rarely reported. Some of the most valuable learning comes precisely from understanding why molecules failed — whether due to synthesis challenges, purification issues, biological inactivity, or toxicity concerns.
That is why more organizations are investing in generating their own comprehensive datasets, including both successful and unsuccessful results. This richer understanding is helping create more informed design cycles and improving the quality of decisions made at each subsequent step.
This is exactly where the transformation starts to make sense. When the DMTA cycle is connected through automation platforms and powered by artificial intelligence models, each step begins talking to the others in real time. Data from the testing phase immediately informs the design of the next molecule. Analysis stops being a final report and becomes a continuous decision engine. This completely changes the speed and quality of choices made inside the lab, and the impact on real lives can be enormous.
How artificial intelligence is changing data analysis in the lab
Wenshu Xu got straight to the point when talking about the role of artificial intelligence inside the modern lab. For her, the greatest value of AI is not in replacing the scientist but in expanding the human capacity to process and interpret massive volumes of data in a timeframe that would be impossible manually. The data analysis generated in a single day of automated experiments can reach millions of data points. Without AI tools, much of that knowledge would simply be lost or stuck in spreadsheets that nobody has time to review with enough depth.
In practice, artificial intelligence models applied to the DMTA cycle work as intelligent, predictive filters. They can:
- Identify patterns in molecular data that would escape even the most trained eyes.
- Suggest which compounds have the highest probability of success in the next rounds of testing.
- Eliminate low-potential candidates before the lab wastes time and resources synthesizing them.
This represents a radical shift in the logic of the process, moving from a nearly random search to a far more targeted and efficient exploration of the available chemical space.
Where AI already shows real value and where it still needs to mature
Wenshu Xu pointed out that one of the areas where AI has already demonstrated significant value is image analysis. The ability to process and interpret large, complex datasets is changing how researchers extract insights from biological experiments.
However, AI still has important limitations. It cannot reliably replace scientific reasoning and domain expertise. Without sufficient context, constraints, and mechanistic understanding, AI systems can generate misleading conclusions. That is why combining AI with physics-based and biology-based models remains so important.
Beyond initial discovery, AI is also starting to influence clinical development. Applications like patient stratification, clinical trial recruitment optimization, and analysis of organoid-derived datasets can significantly improve clinical success rates. A better understanding of toxicity, ADME, PK/PD, and efficacy using more human-relevant models can ultimately reduce costs and improve outcomes across the entire development pipeline.
But Jon Wingfield made a point of highlighting something many people overlook when talking about AI applied to science: the quality of input data is absolutely critical. Artificial intelligence models are only as good as the data that feeds them. If the collection process is inconsistent, if experiments are not reproducible, or if metadata is incomplete, the model will generate predictions that are worthless in practice.
Wenshu Xu reinforced this point by stating that building robust foundation models requires large and reliable datasets that include not just experimental results but also metadata about the experimental environment. This need is pushing the industry toward new methods of data generation at scale. Traditional manual workflows can hardly provide the volume and quality of information required. Technologies like microarrays, microfluidics, and ultra-high-throughput screening will play an essential role in generating the datasets needed to fully realize the potential of AI.
That is why automation is not just a matter of speed. It is also a matter of standardization and reliability of the data that will train and guide AI systems. 🤖
Automation beyond speed: what really changes day to day
When most people hear the word automation, they picture robots doing repetitive tasks faster than humans. And yes, that is part of the story. But within the context of drug discovery, automation plays a far more strategic role than simply speeding up production pipelines.
It creates the conditions for artificial intelligence to work with quality data, allows experiments to run in parallel at scales that were previously impossible, and frees scientists to focus on the decisions that truly require creative reasoning and human expertise.
From repetitive tasks to autonomous systems
Wenshu Xu explained that automation is evolving from simple task execution toward autonomy. Much like the progression from autopilot to self-driving vehicles, labs are moving beyond systems that merely perform repetitive tasks to systems that can diagnose problems, adapt, and potentially self-correct.
Jon Wingfield added that over the past three decades, automation has evolved from simply moving liquids into microplates toward a much broader concept of automated processes. Early efforts focused mainly on increasing assay throughput, moving from 96-well formats to 384-well and beyond. Today, the focus is far more holistic.
A major challenge that remains, according to Wenshu, is more cultural than technical. Scientists often step in when workflows hit problems because manual intervention feels faster in the short term. However, truly autonomous labs require robust systems that can monitor themselves, preserve valuable samples, recover from errors, and keep operating with minimal human involvement.
TTP is a concrete example of how this integration can work. The company has developed lab automation platforms that not only execute experiments but also collect, structure, and make generated data available in a standardized and accessible format for the data analysis and AI systems running in parallel.
At the same time, automation is increasingly viewed as an end-to-end process rather than a collection of individual instruments. Modern systems integrate logistics, experimentation, data capture, analysis, and decision-making into a continuous workflow. This allows teams to focus more on scientific strategy and less on repetitive operational tasks.
The impact on scientific reproducibility
Another important point raised by the experts is the impact of automation on scientific reproducibility — a chronic problem in the pharmaceutical space. When experiments are done manually, inevitable human variations introduce noise into the data. With automated and standardized systems, reproducibility increases significantly, which makes the data more reliable and artificial intelligence models more accurate.
This cascading effect is often underestimated, but it is one of the pillars supporting the entire promise of accelerating drug discovery with technology. 💡
Integrating chemistry and biology: the next big step
A fascinating aspect raised by Jon Wingfield concerns the barriers that still separate chemistry and biology within the drug discovery process. Historically, these two disciplines operated as distinct functions, sometimes even separated geographically as organizations optimized for cost. Today, speed has become far more important than cost alone.
Bringing chemistry closer to biology will allow compounds to move directly from synthesis to biological testing, drastically reducing turnaround times. The labs of the future may evolve into integrated DMTA environments rather than separate chemistry and biology facilities.
Alongside this physical integration, advanced analytics will become increasingly important. New image analysis capabilities and complex cellular models are producing richer datasets than ever before. Organizations that can effectively extract insights from these data streams will be positioned to make better decisions, reduce cycle times, and increase the likelihood of success.
The role of TTP in this transformation
Within this rapidly evolving ecosystem, TTP acts as a bridge between disciplines, technologies, and workflows. According to Jon Wingfield, one of the industry’s biggest challenges is integrating chemistry, biology, automation, and data science into a continuous and efficient discovery engine.
Many organizations have strong capabilities in individual areas but struggle to connect them effectively. This often requires custom instrumentation, tailored automation, and specialized workflow design. TTP helps tackle these challenges by developing customized solutions that connect previously separate processes and enable more efficient DMTA cycles.
Wenshu Xu added that this work also extends to enabling the physical infrastructure behind AI-driven drug discovery. While AI can generate predictions, those predictions need to be validated experimentally. TTP’s expertise in high-throughput experimentation and autonomous lab systems helps create the physical AI ecosystem needed to generate high-quality data and continuously improve discovery outcomes.
The future that is already happening
One of the most interesting takeaways from this conversation with the TTP experts was realizing that much of what sounds like a futuristic trend is already an operational reality in labs around the world. Large pharmaceutical companies and biotech startups are investing heavily in the integration of automation, artificial intelligence, and the DMTA cycle. Early results already show significant reductions in the time needed to identify promising drug candidates, which could mean years shaved off the journey from the lab bench to the patient.
For Jon Wingfield, the biggest challenge now is not technological. It is cultural and organizational. The tools already exist. What still needs to evolve is how multidisciplinary teams work together, how data is shared across departments, and how companies build structures that allow data analysis generated by AI to be effectively incorporated into strategic drug development decisions. This requires not just technology but a pretty profound mindset shift within organizations operating in this space.
There is also the matter of efficiency pressure facing the entire industry. The time and cost required to move a molecule from discovery to the clinic remain extraordinarily high. There is enormous pressure to shorten timelines and make better decisions earlier. Organizations are increasingly adopting automation, advanced analytics, and data-driven approaches to identify promising programs sooner and kill weaker projects before they consume too many resources.
Wenshu Xu wrapped up with a perspective that sums up the current moment nicely: artificial intelligence and automation are not a magic bullet for the complex problems of drug discovery, but they are, without a doubt, the most powerful tools science has ever had available to tackle these problems.
The connected, data-driven, AI-accelerated DMTA cycle represents a new era for the pharmaceutical industry, and its benefits extend far beyond the lab. They ultimately reach the people who need better, faster, and more affordable treatments. 🌍
