Modern drug discovery is built on a paradox. To learn how living systems respond to intervention, we routinely destroy them. Cells are lysed, fixed, stained, sequenced, averaged, and reconstructed after the fact. This gives us a snapshot of what the cells were like before and after, but tells us very little about what happened in between.
This workflow has become so normalized that its limitations are often treated as facts of life. Drug resistance seems sudden and unexpected, while mechanisms of action can take months to untangle. But these are direct consequences of how we measure biology.
A growing group of researchers and technologists are beginning to question this methodology and instead measure biology in real time. In a recent Nature Communications study, researchers used densely sampled, time-resolved measurements to follow cancer cells as they were exposed to targeted therapies. Rather than comparing only baseline and endpoint states, they tracked how gene expression programs and chromatin states evolved across many intermediate timepoints.
What emerged was not a smooth transition from sensitive to resistant cells, but a structured progression through intermediate cellular states. Importantly, cells did not drift randomly into resistance; they passed through predictable transitional states in which early molecular changes foreshadowed later survival outcomes. These intermediate cellular states could be prime targets for intervention before full therapeutic resistance is established.
This type and depth of information is crucial to better understanding disease processes, because it moves the focus away from static classification and toward the dynamics of how systems evolve under pressure. Rather than asking whether a cell is sensitive or resistant, the question becomes how it transitions between these states, and what molecular events govern that transition.
Biology is not static
However, while this approach is much more informative than traditional endpoint assays, it still has limitations. Each timepoint in these studies is still generated by destroying a population of cells, extracting molecular information, and reconstructing dynamics afterwards. What appears as a continuous trajectory is, in practice, a carefully stitched approximation built from discrete, non-overlapping snapshots.
This distinction matters because the most informative biology may occur between the sampled moments. Early metabolic shifts, transient stress responses, and rapidly reversible states can be missed entirely if they do not align with the chosen sampling schedule. In other words, increasing the number of timepoints improves resolution, but does not fundamentally change the fact that the system is still being observed indirectly.
This gap between observation and process has led to interest in methods that can remain continuously coupled to living systems, capturing their state without interrupting or resetting it. Rather than treating cells as endpoints to be repeatedly sampled, these approaches aim to treat them as continuously evolving systems that can be measured in place, preserving both their temporal structure and their internal heterogeneity as they respond to perturbation.
DDN spoke with Parmita Mishra, founder and CEO of Precigenetics, who is developing a label-free optical platform designed to interrogate living cells continuously at the level of their native chemistry. Instead of relying on fluorescent tags or destructive readouts, the system uses optical signatures of molecular bonds to track changes in proteins, lipids, and metabolites over time within the same intact cells.
“The idea is that we use subcellular vibrational spectroscopy — of which Raman imaging is a leading example,” she said. “It’s label-free and already well established in materials science, and even used on Mars rovers to detect chemical signatures that could indicate microbial life. The key point is that no fluorescent labels or probes are involved — nothing that perturbs the system you’re trying to measure.”
To support this approach, the company has developed custom hardware and in-house microfluidic chips for live-cell Raman measurements. By keeping cells in tightly controlled conditions with physiological temperature and five percent CO₂ delivered through microfluidics, the company can continuously collect full-spectrum chemical data from living cells.
Using the same nanometer-scale optical precision found in semiconductor imaging, Mishra said that the company is aiming to create a scalable platform that “will scale everything in our understanding of the cell.”
Measuring the entire cell
What distinguishes this approach from many existing live-cell assays is not simply the ability to observe biology in real time, but the type of information being captured. Most current platforms focus on a limited number of predefined targets or rely on fluorescent labels to track specific pathways. Even in advanced systems, the measurement is often still anchored to a single biological question per experiment.
“This matters because, in drug discovery, the target still dominates most thinking,” said Mishra. “Of course it matters — but in preclinical settings, the target is often wrong or at least incomplete. The issue is that we’re not measuring the rest of the cell.”
By contrast, a chemical readout of the entire cell allows multiple layers of biology to be observed simultaneously. Instead of asking whether a specific protein is activated or inhibited, the system captures coordinated changes across lipids, metabolites, and proteins as the cell responds to perturbation. This makes it possible to see not just whether a drug is working, but how the entire cellular state is reorganizing in response.
This approach reveals dynamics that are often invisible in conventional assays. A cell that appears to be dying at an early timepoint may later recover through compensatory metabolic shifts. Conversely, subtle biochemical changes that precede visible phenotypic effects may signal the earliest stages of resistance or toxicity long before they would be detected through standard viability or sequencing-based approaches.
Cells are data
The Precigenetics platform is designed to generate continuous, high-dimensional readouts of living cells over time. “For us, we’re generating about two to three gigabytes of data per cell, every single cell, every two hours,” Mishra said.
Rather than a single endpoint measurement, each experiment is a dense time-resolved record of how a cell’s internal chemistry evolves under perturbation. At scale, this generates large volumes of structured data that, in principle, could be used to train predictive models of cellular behavior.
Mishra frames this as part of a broader gap in the field: the lack of a shared, high-quality reference layer for cellular state and dynamics. “We are never going to solve biological problems without a protein database equivalent for these companies to use,” she said.
To address this, the company is building what it calls Cleopatra, an AI-based toxicity model. “We basically want to train Cleopatra and ideally release it similar to how AlphaFold was released — making it broadly accessible to the community,” she said.
Measuring the continuous
About nine out of ten drug candidates still fail in clinical trials, with toxicity alone accounting for roughly 30 percent of those failures. Despite advances in sequencing, imaging, and computational modeling, most of the field is still operating on a fragmented view of biology, rather than a continuous understanding of how living systems respond over time.
There is growing recognition that dynamic biology requires dynamic measurement. Whether through densely sampled time-series omics, advanced imaging approaches, or new label-free optical systems, the common thread is a move away from destructive, endpoint-driven assays toward methods that remain coupled to living systems as they evolve.
The implication of this shift is not simply better resolution, but better inference. As measurement becomes more continuous and more complete, the hope is that biology itself becomes more legible — translating into more reliable therapeutics, fewer late-stage failures, and a drug discovery process that is finally aligned with the dynamics of the systems it is trying to intervene in.
