DDW Editor Reece Armstrong speaks to Yerem Yeghiazarians, CEO and Co-Founder of Soley Therapeutics, about the limitations of current assays and how the company’s live cell approach can improve drug discovery.  

RA: What are the current limitations of static assays and single endpoints in assessing how cells respond to drugs?

YY: Most conventional assays provide a static snapshot of a process that is inherently dynamic. Cells continuously sense their environment and adjust their behaviour in response to stress, nutrients, and perturbations such as drugs. When readouts rely on fixation, lysis, or other destructive steps, the temporal sequence of events that determines outcome is lost. 

Single endpoints also require committing in advance to what is assumed to matter, whether that is viability, a reporter, or a specific pathway. Those approaches implicitly assume that biology can be reduced to a fixed endpoint, whether that is target engagement, pathway modulation, or cell viability. In reality, cells are living systems that continuously sense stress, adapt, and make decisions about survival or death. 

At Soley, the foundational insight is that many complex diseases are not driven by a single broken node, but by how cells process stress over time. Static assays miss that entirely. They often tell what happened, but not how or why it happened. As a result, discovery programmes can advance compounds that look promising against a predefined endpoint, but fail to meaningfully reshape the cellular behaviour that actually matters in disease. 

RA: How can ‘live cell dynamics’ be applied to improve drug discovery?
 

YY: Live cell dynamics allows biology to be observed as a process rather than inferred as an outcome. That shift is central to Soley’s approach. Instead of starting with a target and asking whether a compound binds or inhibits it, discovery begins by observing how living human cells respond to perturbation over time and how stress signals propagate through the system. 

This makes it possible to identify compounds that meaningfully alter cell fate trajectories, not just those that trigger a desired endpoint at a single timepoint. In practice, this leads to earlier signal detection, better separation of selective pharmacology from nonspecific stress responses, and a clearer understanding of which candidates are reshaping cellular behaviour in a way that is consistent with durable therapeutic benefit. 

RA: Could you summarise the how live cell dynamics works?
 

YY: Live cell dynamics begins with continuous, label-free imaging of living human cells across multiple doses and timepoints, using brightfield microscopy to preserve native cellular behaviour without introducing bias from predefined markers. A self-supervised machine learning framework learns informative representations directly from the images, rather than being told what to look for.  

The model is trained to be robust to the specific challenges of brightfield imaging, including low contrast and technical batch effects, and to extract subtle, time-dependent patterns that reflect underlying cell state. Those learned representations are then used to quantify phenotypic activity and compare responses across compounds. 

The result is a representation of cell state that captures how information flows through the cell in response to stress and drug exposure. This framework allows dynamic cell behaviour to be quantified, compared, and interpreted at scale. 

RA: What mechanistic information can this framework reveal? 

YY: The most important mechanistic insight is not tied to a single pathway, but to how cells transition between states. At Soley, mechanism is understood in terms of cellular trajectories, whether a compound pushes stressed cells toward recovery, adaptation, quiescence, or death. 

By analysing responses across dose and time, the framework improves mechanism of action classification and can surface polypharmacology that would be difficult to detect using static assays. Importantly, mechanistic insight emerges from behaviour first, with target and pathway understanding layered in deliberately once the biology proves compelling. This reverses the traditional order of discovery while remaining rigorous and grounded in data. 

RA: Can this method be extended to other areas of biology? 

YY: Yes. The broader opportunity extends well beyond any single indication because Soley’s platform is built around conserved stress-sensing biology that operates across tissues and diseases. The central question is not simply whether a cell is alive or dead, but how it processes stress and what trajectory that stress response puts the cell on. 

In oncology, the platform is used to selectively increase stress in cancer cells, pushing already vulnerable cells past a threshold toward death while sparing healthy cells. In other disease areas, the same framework can be applied in the opposite direction, identifying compounds that reduce or rebalance stress and restore cells toward healthier, more functional states. This is particularly relevant in areas such as neurodegeneration and metabolic disease, where chronic stress drives dysfunction rather than uncontrolled proliferation. 

Because the platform learns fundamental patterns of live cell behaviour and stress response, it can be applied across biological contexts without redefining the system for each disease. That ability to modulate stress in different directions, using the same underlying framework, is what allows Soley to build a growing, diversified pipeline rather than a collection of isolated programmes. 

RA: Can this method be scaled to improve costs in drug development? 

YY: Scalability was a design requirement from the start. Brightfield imaging is reagent-free, cost-effective, and well suited to automation. Combined with Soley’s internally built infrastructure, this enables high-throughput, information-rich experimentation without the escalating costs associated with complex assay stacks. 

More importantly, richer early data reduce downstream attrition. By identifying poor candidates earlier and advancing compounds with coherent, biologically grounded profiles, the platform improves capital efficiency across the development lifecycle. Cost reduction is not just about cheaper assays, but about better decisions made sooner.

RA: How important have advances in machine learning and AI in recent years been for enabling this type of project? 

YY: Recent advances in machine learning, particularly self-supervised learning, have been critical. Earlier approaches required extensive manual labelling or predefined features, which limited scalability and introduced bias. Self-supervised methods make it possible to learn directly from raw data, capturing structure and variation that would otherwise be inaccessible. 

Equally important are advances in model architectures and training strategies that allow meaningful biological signals to be separated from technical noise in large, heterogeneous datasets. When combined with decades of biological insight into stress sensing and cell fate, these tools allow questions that were previously unapproachable to be addressed systematically. 

RA: Could you discuss the usage of self-supervised machine learning and how your framework overcomes existing challenges when using brightfield microscopy? 

YY: Brightfield microscopy has long been attractive because it is non-destructive and scalable, but difficult to analyse due to low contrast and technical variability. The framework addresses these challenges through training strategies that explicitly separate biological signals from technical noise, including batch-aware learning and augmentation approaches that account for focal variation. 

Self-supervised learning allows the model to learn directly from the data, rather than relying on manual annotations or fluorescent proxies. This makes it possible to extract subtle, time-dependent phenotypes that reflect true biological response rather than experimental artefacts. 

RA: How are you utilising this method in your own work to advance in-house pipelines? 

YY: Live cell dynamics is one component of Soley’s broader platform and is used to capture how living human cells respond to compounds over time. It provides a dynamic, label-free view of cell state that helps inform early discovery decisions, particularly in programmes that begin without a predefined target. 

As compounds advance, Live cell dynamics data feed into SoleyMaps, which serve as a shared reference framework of stress response patterns. This allows candidates to be compared and prioritised based on how they move through defined state spaces, rather than on single endpoints, helping identify molecules that combine meaningful biology with developability. 

Later in the pipeline, these insights support lead optimisation by guiding medicinal chemistry toward shaping cellular behaviour, not just improving potency. Target and mechanism elucidation are then pursued deliberately, once a compound has demonstrated the right biological profile. 

Using this integrated approach, Soley has built a growing in-house pipeline, led by oncology programmes advancing toward the clinic, alongside additional programmes in non-oncology areas where restoring or reshaping cell state may provide therapeutic benefit.

Biography

 Yerem Yeghiazarians is Co-Founder and Chief Executive Officer of Soley Therapeutics. A clinician scientist with more than two decades of experience in cardiology, stem cell biology, and translational research, he leads Soley’s biology-first discovery strategy and guides the company’s evolution from foundational science into a clinical-stage organisation.  

From DDW Volume 27 – Issue 2, Spring 2026 – Read the digital issue here