The work, led by Alexandre Coudray, Romain Forey, and colleagues at EPFL and collaborating institutions, introduces two linked algorithms named CLIMB and CLIFF. CLIMB addresses the classical problem of deconvolution: given a bulk RNA-sequencing profile that represents the summed expression of thousands of cells, estimate both the proportions of each cell type present and the expression programs those cell types are running. CLIFF then builds on that decomposition to do something more ambitious: it resolves bulk drug sensitivity measurements, which normally report only an averaged response across an entire sample, into cell-subtype-specific and single-cell-specific estimates of how sensitive each cellular component is to a given compound.

The conceptual foundation of the approach rests on a mapping between two data worlds that rarely meet. Single-cell RNA sequencing provides exquisitely detailed snapshots of individual cells, but it is expensive, technically demanding, and difficult to pair with functional readouts such as drug screening. Bulk RNA sequencing is cheap, abundant, and routinely paired with drug-response data in large clinical cohorts, but it lacks cellular resolution. CLIMB and CLIFF act as a bridge between these worlds: they use single-cell reference data to teach a model what the expression signatures of specific cell types look like, and then apply that learned mapping to interpret bulk profiles. The authors describe this as single-cell-to-bulk transcriptome mapping, and it is this transfer of resolution from the single-cell domain to the bulk domain that gives the framework its power.

Validation began in a controlled setting. The team mixed leukemic cell lines together in the laboratory in known proportions and generated bulk profiles of these artificial mixtures. Because the true cellular composition was experimentally established, the researchers could measure exactly how well CLIMB recovered the ground truth. The algorithm performed accurately, and the same mixtures, which had also been profiled for their responses to drugs, allowed the team to verify that CLIFF could correctly attribute drug sensitivity to the constituent cell lines.

Real-world applications, however, are messier than laboratory mixtures. In clinical practice, the single-cell reference data used to train a deconvolution model rarely comes from the same patients, the same sequencing platforms, or even the same disease context as the bulk samples being analyzed. To test whether CLIMB could withstand these so-called reference-to-target mismatches, the researchers constructed large-scale pseudo-bulk analyses: they aggregated single-cell RNA-sequencing data from independent references across two disease contexts and five cancer types, deliberately introducing the kinds of discrepancies that arise in real studies. Even under these challenging conditions, CLIMB accurately inferred cellular composition and outperformed state-of-the-art bulk deconvolution methods, suggesting that the framework is robust to the imperfections of real reference data.

One of the most striking demonstrations of the method comes from acute myeloid leukemia, or AML, an aggressive blood cancer in which a rare population of leukemic stem cells is thought to sustain the disease and drive relapse after treatment. These cells are notoriously difficult to study because they are scarce and hard to isolate, yet they are believed to be a critical target for curative therapy. CLIMB proved capable of estimating leukemic stem cell fractions with high accuracy across three test datasets and three independent patient cohorts, offering researchers a computational window into a cell population that has long evaded routine measurement in bulk clinical samples.

With the deconvolution engine validated, the team turned to the question that motivates much of precision oncology: which drugs actually kill which cells? To answer it at scale, they applied their framework to the beatAML cohort, currently the most extensive database of ex-vivo drug screening data, in which primary AML samples from patients have been tested against hundreds of compounds while their molecular profiles were simultaneously recorded. Running CLIFF on this resource generated a cell-subtype-resolved drug sensitivity landscape covering 121 compounds, effectively transforming a database of averaged responses into a map of which drugs act on which cellular populations within the leukemia.

The analysis surfaced several compounds that appear putatively active against leukemic stem cells, a finding with direct therapeutic implications. Among the candidates identified were midostaurin, a multikinase inhibitor already approved for use in AML; A-674563, an AKT inhibitor; SNS-032, an inhibitor of cyclin-dependent kinases; and foretinib, another multikinase inhibitor. Because leukemic stem cells are implicated in treatment resistance and relapse, compounds that preferentially target them could inform new therapeutic strategies, particularly for patients whose disease persists after standard induction chemotherapy. The authors emphasize that these are computational predictions that would require experimental and clinical validation, but the framework provides a systematic way to prioritize candidates that would otherwise be hidden inside bulk averages.

Beyond the specific findings in leukemia, the study articulates a broader methodological vision. Bulk functional measurements, whether drug screens, CRISPR perturbation screens, or other phenotypic assays, are abundant in biomedicine precisely because they are scalable and inexpensive. The CLIMB-CLIFF framework offers a general recipe for endowing such measurements with cellular resolution by leveraging the growing library of single-cell atlases. The authors position the approach as an attractive new computational tool for drug development and precision medicine, one that could in principle be extended beyond oncology to any context where heterogeneous cell mixtures obscure the signals that matter, from autoimmune disease to neurodegeneration.

The work also arrives at a moment of intense interest in computational deconvolution, as the field grapples with how to reconcile the explosion of single-cell data with the continued dominance of bulk assays in clinical and pharmaceutical settings. By demonstrating accuracy across controlled mixtures, cross-dataset pseudo-bulk benchmarks, multiple cancer types, and a large clinical drug-screening cohort, the EPFL-led team has provided one of the more comprehensive validations of the deconvolution concept to date. If the predictions hold up in laboratory follow-up, the ability to computationally resolve which cells in a tumor respond to which drugs could reshape how drug screens are interpreted, how patient stratification is performed, and ultimately how therapies are matched to the specific cellular vulnerabilities hidden within each patient’s disease.

Subject of Research: Computational deconvolution of bulk transcriptomic and drug sensitivity data to single-cell resolution

Article Title: Computational deconvolution of drug sensitivity via single-cell–to–bulk transcriptome mapping

Article References: Coudray, A., Forey, R., Bejar Haro, B., Martins, F., Carlevaro-Fita, J., Sheppard, S., Offner, S. E., La Manno, G., Obozinski, G., & Trono, D. (2026). Computational deconvolution of drug sensitivity via single-cell–to–bulk transcriptome mapping. PLOS Computational Biology, 22(10), e1014862. https://doi.org/10.1371/journal.pcbi.1014862

Image Credits: AI Generated

DOI: 10.1371/journal.pcbi.1014862

Keywords: computational biology, deconvolution, single-cell RNA sequencing, bulk RNA-seq, drug sensitivity, acute myeloid leukemia, leukemic stem cells, machine learning, beatAML, precision medicine, drug development, bioinformatics

The scale of the analysis is what sets the study apart. The research team, led by Dr. Glen Pridham and Yoav Hayut from Alon’s lab in Weizmann’s Molecular Cell Biology Department, drew on two enormous datasets: results from the US National Health and Nutrition Examination Survey, collected over more than two decades, and medical records spanning roughly 20 years from Clalit, Israel’s largest health insurer. Datasets of this size are rarely brought to bear on menopause research, and the reason is simple: they usually lack the one piece of information that matters most for this kind of analysis, namely the age at which each woman had her final period.

That missing information creates a fundamental problem for anyone trying to pin down what happens at menopause itself. If women reach menopause at different ages, and a researcher aligns their medical records by calendar age, the true signal of the transition gets smeared across time. A sharp change that occurs at age 51 in one woman and age 54 in another looks, in the averaged data, like a gradual change stretched over several years. The apparent slowness of menopause, in other words, could partly be an artifact of the way the data are pooled rather than a property of the biology itself.

To cut through this blur, the Weizmann team borrowed an idea from an unexpected field: astronomy. Astronomers routinely sharpen images distorted by the optics of their telescopes by mathematically correcting for those distortions, a process known as deconvolution. The researchers developed a computational method inspired by this approach that removed the blurring introduced by differences in the age at which women reach menopause. Once that correction was applied, what emerged from the data was not a gentle slope but a distinct physiological turning point, a sudden shift in dozens of bodily measures concentrated around the time of the last menstrual period.

The breadth of the changes came as a surprise. Of the 90 types of laboratory tests examined, 52 showed abrupt changes at the transition, covering nearly all body systems. The affected measures included markers of bone health, liver function, kidney function, and cholesterol levels, alongside the expected shifts in sex hormones. This suggests that menopause is not confined to the reproductive axis but registers across the organism, with the end of ovarian hormone production rippling outward into metabolism, skeletal maintenance, and organ function in a coordinated and comparatively rapid way.

Two comparison groups in the data helped sharpen the interpretation. No such abrupt changes were observed in men of similar ages, indicating that the turning point is specific to the female transition rather than a general feature of aging in midlife. And among women receiving menopausal hormone therapy, the changes were less pronounced, a pattern consistent with the idea that the declining hormone levels themselves drive much of the body-wide shift. While the study’s design does not by itself establish hormone therapy as protective, the contrast offers a natural experiment hinting that supplementing the hormones lost at menopause may soften the physiological transition.

The timing of the underlying hormonal changes also proved more dramatic than expected. The sudden shifts at menopause were preceded by accelerating changes in sex hormone levels that began more than a decade before the final period, several years earlier than previous research had suggested. In practical terms, the body appears to enter a period of accelerating hormonal change in the mid-to-late thirties or early forties for many women, long before the first obvious symptoms of the transition appear. The final event at menopause, on this view, is the endpoint of a long buildup rather than the beginning of the story.

Because most of the abrupt changes identified in the study are associated with increased disease risk, the findings carry a clear message for preventive medicine. Bone loss, adverse changes in cholesterol, and declining kidney and liver measures are all modifiable risk factors, and interventions such as exercise and dietary improvement are most effective when they start early. Our study extends the window for potential preventive care to almost a decade before menopause, Alon says. That reframing turns the decade before the last period from a quiet prelude into a critical window in which women and their physicians can act on measurable, trackable changes in blood tests.

The methodological contribution may prove as consequential as the biological one. The deconvolution-style approach the team developed is not specific to menopause; it can, in principle, be applied to any large medical dataset in which a key event happens at different ages for different people, allowing researchers to recover the true sharpness of the underlying transition. More broadly, the method we developed may help study other major changes in life, such as puberty, pregnancy or the onset of disease, Alon notes. If other life transitions that look gradual in population data turn out to be sharp once the blur is removed, the study could reshape how medicine thinks about aging and disease onset more generally.

The study also involved Dr. Noa Lavi-Shoseyov and Prof. Michal Neeman of Weizmann’s Immunology and Regenerative Biology Department, Dr. Noa Hovav of the Menopause Community Center at Clalit Health Care, and Dr. Yoel Toledano of the Helen Schneider Women’s Hospital at Rabin Medical Center, with Dr. Lavi-Shoseyov additionally affiliated with the IVF Unit in the Department of Obstetrics and Gynecology at Kaplan Medical Center. Taken together, the work recasts menopause as a discrete physiological event with a long and measurable runway leading up to it. For the more than one million women whose records made the discovery possible, and for the millions more approaching the transition, the practical takeaway is that the years before menopause matter far more than has been assumed, and that routine laboratory tests, viewed through the right computational lens, can reveal a turning point in time to act on it.

Subject of Research: A computational analysis of large medical datasets revealing a sudden body-wide physiological transition at menopause

Article Title: Menopause: more than a slow change, it’s a sudden shift across the body

Article References: Menopause: more than a slow change, it’s a sudden shift across the body. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: menopause, Weizmann Institute, Nature Medicine, deconvolution, sex hormones, preventive medicine, bone health, cholesterol, hormone therapy, NHANES, Clalit, biomarkers

The traditional method is elegantly simple. Researchers count the density of tracks crossing a polished mineral surface, measure the uranium concentration, and combine these with the decay constant in a standard age equation. For minerals that cooled quickly, such as ash or volcanic extrusions, this produces a precise crystallisation age. But the equation is also routinely applied to apatite that cooled slowly through the so-called partial annealing window, the temperature range between roughly 120 and 60 degrees Celsius in which tracks form but are gradually shortened and erased. In that regime the resulting single apparent age is at best a rough average of a complicated thermal story, and Jensen argues it deserves far more scrutiny.

The key insight behind the new work is that track length carries information that the conventional equation ignores. Fission tracks are born at a standard initial length of about 16 micrometres and then shorten over time as heat allows the crystal to heal itself, a process called annealing. Crucially, the oldest tracks in a grain have generally experienced the most annealing and are therefore the shortest, while the youngest tracks remain the longest. In principle, then, a track of a given length can be dated simply by counting how many shorter tracks exist below it, adding one, and dividing by the rate at which tracks are generated per unit volume of crystal. That deceptively clean relationship was established theoretically decades ago, but turning it into a practical dating tool has proved difficult.

The difficulty lies in blurring. Tracks of the same true age do not all have the same length. The fission fragments carry a spread of kinetic energies, the crystal lattice shortens tracks at different rates depending on their orientation relative to the crystallographic c-axis, and the etching and measurement procedures used to make tracks visible under the microscope introduce further biases. The result is that each column of a measured track-length histogram mixes tracks from several different time intervals, and the tidy one-to-one link between length and age is smeared out. Without correcting for this smearing, any attempt to assign ages to individual length classes would be meaningless.

Jensen’s solution borrows a mathematical technique from signal processing: deconvolution, the same family of methods used to strip noise from telecommunications signals. The idea is that the measured histogram can be treated as a weighted sum of reference histograms, each representing tracks generated in a short time interval and then annealed to a particular degree. Those reference distributions, called kernels, come from laboratory experiments in which apatite samples are irradiated with neutrons in a reactor and then heated at controlled temperatures and durations, producing known track-length spreads. If the measured histogram can be decomposed into the right weighted combination of these kernels, the weights reveal how many tracks were produced in each time interval, restoring the link between histogram columns and time.

The new study advances this approach in two important ways. First, it applies the deconvolution to c-axis projected tracks, a standardised measurement protocol in which track lengths are corrected along the crystal’s principal axis. Because tracks of the same age tend to cluster along an elliptical curve in the length-versus-angle diagram, projecting them onto the c-axis already gathers contemporaneous tracks together, acting as a partial deconvolution in itself. Second, the inversion is formulated in a fully probabilistic framework based on the work of geophysicist Albert Tarantola, using an extended least-squares method that incorporates both the uncertainties of the measurements and prior information about the expected answer. Because the total number of counted tracks is fixed, the histogram columns are strongly covariant, and the method handles this explicitly through variance-covariance matrices rather than treating each bin as independent.

One stubborn mathematical problem is that least-squares inversion can return negative numbers of tracks, which is physically impossible. Jensen tames this by choosing positive prior values with variances large enough that the priors guide the solution without dominating the data, ensuring that the deconvolved histogram columns remain mainly positive. The variances are also chosen generously enough that, in principle, all tracks could end up in a single bin, so the prior does not impose an artificially rigid shape on the result. The full calculation, implemented in the open-source Octave platform and released with the paper, converts the deblurred length histogram into a histogram of time intervals and then cumulates those intervals from the youngest tracks backwards to assign an age to the oldest track in each column.

To test the method, Jensen applied it to twenty apatite samples from Ellesmere Island and Northwest Greenland, part of a large dataset assembled to reconstruct the Phanerozoic tectonic history of the Arctic. The results are striking. In a Late Devonian sandstone sample, the deconvolution extracted three statistically distinct age nodes at roughly 134, 274 and 443 million years, with non-overlapping uncertainties. The oldest of these predates the deposition of the sandstone itself, revealing that the apatite grains carried inherited tracks from earlier mountain-building episodes before they were eroded and redeposited. By comparing track ages with the known depositional age, the method could even separate post-depositional tracks from inherited ones, showing in this case that about 70 percent of the tracks formed after burial.

The deconvolution also sharpens features that are nearly invisible in raw data. For several samples, histograms that appeared broad and unimodal in the original measurements revealed clear bimodality after deblurring, a signature of distinct thermal episodes. One sample told a three-act story: a long stretch between about 193 and 90 million years ago when maximum temperatures reached the upper part of the partial annealing window, a shorter cooling phase through the lower part of the window, and then a prolonged period of moderate temperatures below 60 degrees Celsius lasting to the present. The method even proved useful as a quality check, flagging samples whose histograms showed unrealistically sharp peaks caused by too few measured tracks, since the forward calculation from the deconvolved histogram could not reproduce such features.

The broader payoff is a new layer of information for reconstructing how mountains rose, basins filled and plates shifted. Across the sample suite, the oldest track ages typically came out about 20 percent older than the conventional central ages, and they broadly align with the timing of the last thermal peaks above 100 degrees Celsius inferred from independent thermal history modelling, albeit with large uncertainties. Most samples yielded between two and five reliable age nodes, averaging three, from a single histogram. Jensen suggests these multiple age estimates could guide the choice of age-temperature constraints in thermal history simulation models, potentially tightening reconstructions of past temperatures and tectonic events. For a technique that has spent half a century squeezing a single number out of each mineral grain, the prospect of reading an entire chronological sequence from one crystal marks a quiet but potentially transformative shift in how geologists read Earth’s thermal diary.

Subject of Research: Extraction of multiple thermal history ages from c-axis projected apatite fission tracks using probabilistic deconvolution

Article Title: Extraction of multiple ages from c-axis projected fission tracks

Article References: Jensen, P. K. (2026). Extraction of multiple ages from c -axis projected fission tracks. Geochronology, 8(2), 373-386. https://doi.org/10.5194/gchron-8-373-2026

Image Credits: AI Generated

DOI: 10.5194/gchron-8-373-2026

Keywords: fission track dating, apatite, thermochronology, deconvolution, uranium-238 fission, partial annealing zone, c-axis projection, geochronology, tectonic history, inverse theory, track length distribution, Ellesmere Island

The central challenge the team is tackling is one of scale and specialization. The eye is not a single uniform organ but a collection of highly specialized tissues and cell types, including the retina’s photoreceptors, retinal pigment epithelial cells, corneal cells, lens cells, and vascular tissues, all of which must coordinate precisely to make vision possible. Each of these cell types relies on its own repertoire of transcriptional regulators, the proteins and regulatory elements that determine which genes are switched on or off at any given moment. When these regulatory programs malfunction, the consequences can include degenerative conditions such as age-related macular degeneration, glaucoma, and diabetic retinopathy, which together affect hundreds of millions of people worldwide. Understanding the relationship between tissue-specific transcriptional regulators and visual function, Wang argues, is a cornerstone of vision research.

Wang’s collaborator on the project is Lin Xu, an assistant professor at UT Southwestern Medical Center. The partnership brings together complementary expertise that spans Bayesian statistics, bioinformatics, single-cell sequencing, omics data analysis, ocular biology, and clinical experience. This interdisciplinary approach is increasingly seen as essential in modern genomics, where the raw data generated by sequencing instruments far outstrip the capacity of traditional analytical methods. Bayesian statistical frameworks, which allow researchers to formally quantify uncertainty and integrate prior biological knowledge into their models, are particularly well suited to the noisy, high-dimensional, and often incomplete datasets that characterize contemporary eye research.

The award builds on Wang’s established record of leveraging artificial intelligence and advanced statistical methods to help scientists make sense of complex diseases. Her research program focuses on creating computational tools that uncover important biological patterns hidden within massive datasets, patterns that would be practically impossible for human analysts to detect by inspection. By identifying these patterns, the tools help researchers narrow down which genes, regulatory elements, and molecular pathways deserve the closest scrutiny as potential targets for future study and therapeutic intervention. In an era when a single experiment can generate terabytes of genomic information, the bottleneck in biomedical discovery has shifted from data generation to data interpretation, and that is precisely the bottleneck Wang’s methods are designed to relieve.

To advance the study, the researchers will develop two distinct computational tools, each addressing a different layer of the problem. The first is a scalable analytical engine designed to process very large genomic datasets and identify the regulators that are key to ocular health and disease. Scalability is not a trivial requirement in genomics. Modern studies routinely profile thousands of samples across dozens of tissue types and conditions, and algorithms that perform well on small pilot datasets often collapse under the computational weight of full-scale data. Wang has emphasized that the algorithm needs to run smoothly on very large collections of data, a constraint that shapes everything from the mathematical formulation of the model to the software engineering choices behind its implementation.

The second tool tackles a problem that has long frustrated researchers studying complex organs: signal mixing. Eye tissue samples, like samples from most biological organs, contain many different cell types with distinct functions, and standard bulk sequencing methods measure the average gene activity across all of those cells simultaneously. When researchers analyze a whole tissue sample, the signals from individual cell types are blended together, obscuring the very cell-type-specific regulatory behavior that matters most for understanding disease. The new deconvolution tool will computationally separate those mixed signals, estimating how a particular transcriptional regulator behaves in each constituent cell type without requiring expensive single-cell experiments for every regulator under investigation.

The significance of this second tool is difficult to overstate. Single-cell sequencing technologies, which measure gene expression in individual cells, have revolutionized biology by revealing cellular heterogeneity that bulk methods average away. However, they remain costly, technically demanding, and impractical to deploy for every regulator, tissue, and disease condition a researcher might want to study. A reliable computational deconvolution method offers a middle path: researchers can continue to generate relatively inexpensive bulk data while using the algorithm to infer cell-type-specific regulatory activity, reserving single-cell experiments for the most promising hypotheses. By pinpointing the key regulators that behave abnormally in specific cell populations, the tools will help researchers focus on the most promising targets, saving both time and resources in the search for therapies.

Together, the two tools will allow the team to study the eye at two complementary levels: the overall tissue level, where regulators can be assessed across large collections of samples and disease conditions, and the individual cell-type level within each tissue, where the mechanistic details of health and disease actually play out. Wang has described the long-term goal as improving understanding of eye disease mechanisms and helping the field identify promising directions for future diagnostics and targeted therapies. Targeted therapies, which act on specific molecular pathways rather than broadly suppressing symptoms, represent one of the most promising frontiers in ophthalmology, but they depend on knowing exactly which molecular switches to flip and in which cells. The computational framework under development is intended to provide exactly that kind of mechanistic roadmap.

The National Eye Institute’s support for the project underscores a broader strategic bet within federal vision research: that computational and data-driven approaches will increasingly drive the next generation of discoveries about how the eye works and why it fails. The institute, part of the National Institutes of Health, funds research aimed at preserving and improving vision, and grants of this size typically signal confidence that the proposed methods can yield tools of lasting value to the wider research community. Morteza Khaledi, dean of UT Arlington’s College of Science, called the grant an outstanding achievement and a testament to the strength, significance, and growing impact of Wang’s interdisciplinary research program, reflecting the university’s investment in statistical and data science approaches to biomedical problems.

The project also fits into a larger portfolio of AI-driven biomedical work at UT Arlington. Wang is co-leading a related effort with Junzhou Huang, a professor in the university’s College of Engineering, on a study that combines artificial intelligence and Bayesian learning to accelerate drug design, an area where machine learning methods are already compressing discovery timelines that once stretched over decades. Taken together, these initiatives position the university at the intersection of statistics, computer science, and medicine, where the tools of data science are being forged into instruments of biological discovery. If the new eye-focused tools perform as intended, they could give researchers around the world a clearer view of the molecular circuitry of vision, and a better chance of protecting it when disease threatens.

Subject of Research: AI-driven identification of transcriptional regulators controlling ocular health and eye disease

Article Title: UT Arlington uses AI to uncover drivers of eye disease

Article References: UT Arlington uses AI to uncover drivers of eye disease. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: artificial intelligence, eye disease, National Eye Institute, transcriptional regulators, Bayesian statistics, single-cell sequencing, genomics, bioinformatics, UT Arlington, vision research, targeted therapies, deconvolution

The difficulty lies in a fundamental mismatch between what scientists can measure and what they want to know. Ideally, researchers would track protein concentrations in single cells as they progress through the cycle, but single-cell proteomics currently lacks the sensitivity required for the tiny volumes of yeast cells, and unlike RNA, proteins cannot be amplified to compensate for minute sample amounts. Fluorescence microscopy with tagged proteins offers one alternative, but it suffers from incomplete fluorophore degradation, artefacts from the tags themselves, and limited temporal resolution. The practical compromise has long been to synchronise a population of cells chemically or mechanically and then measure protein abundances across the culture over time using mass spectrometry. The trouble is that synchronisation is never perfect. Cells drift out of alignment, especially in budding yeast, where asymmetric division gives newborn daughters a longer G1 phase than their mothers, and intrinsic biochemical stochasticity progressively scatters the population across cell cycle stages.

This desynchronisation acts like a blur filter, smearing out the true oscillations in protein concentration and systematically underestimating how dynamic the proteome really is. To undo that blur, the Groningen team, led by Andre Zylstra and Matthias Heinemann, turned to computational deconvolution, a mathematical technique that solves the inverse problem of reconstructing an underlying signal from an observed, distorted one. Their approach models each bulk measurement as a weighted mixture of cell cycle stage-specific concentrations, where the weights reflect how the sampled cell volume is distributed across the cycle at the moment of sampling. Expressed as a matrix equation, the relationship between the true single-cell dynamics and the population-averaged measurements is captured by a convolution matrix that encodes the blurring effect of desynchronisation specific to each experiment.

Estimating that convolution matrix accurately was the central technical hurdle. The researchers built a sophisticated computational model that simulates the temporal evolution of a yeast population, cell by cell, tracking each individual cycle through its phases of early G1, late G1, S/G2, anaphase and telophase, along with cell volume growth in G1 and after budding. Crucially, the parameters for these simulations were not invented but measured. Using time-lapse fluorescence microscopy of cells growing in microfluidic devices, the team tracked 192 complete cell cycles, identifying key events such as START, budding, karyokinesis and cytokinesis with the help of fluorescently tagged histone H2A and the cell cycle inhibitor Whi5. From these data they extracted distributions of phase durations, birth volumes and growth rates, fitted as multivariate log-normal distributions that capture the natural variability between cells. Because budding yeast divides asymmetrically, the model treats mother and daughter cycles distinctly, with daughter early-G1 phases lasting a median of sixty minutes compared with just fifteen minutes for mothers.

The model was then fine-tuned to each specific proteomics experiment using independent measurements of cell cycle phase distributions and cell volume distributions taken from the same cultures. When the standard model was compared with experimental populations, simulated cells reached START roughly twenty to fifty minutes earlier than their real counterparts, likely reflecting stress from the elutriation procedure or differences between shake flasks and microfluidic chambers. By adjusting the log-normal parameters with particle swarm optimisation, the team brought simulations into close agreement with the measured populations, producing convolution matrices that faithfully represented the desynchronisation present in each replicate time course. This volume-aware approach marked a significant advance over earlier deconvolution studies, which relied mainly on DNA content or budding index data and largely ignored the substantial influence of cell size on population-averaged concentration measurements.

With the forward model in place, the team confronted the second major obstacle: deconvolution is mathematically ill-conditioned, meaning that even small amounts of noise in the input data can explode into dramatic distortions in the reconstructed trajectories. The researchers demonstrated this vividly with synthetic data, showing that a simple non-negative least squares approach fails catastrophically when a modest amount of Gaussian noise is added to an otherwise perfect signal. Their solution was a regularised least squares algorithm that penalises roughness in the reconstructed concentration profile, favouring smooth solutions consistent with the expectation that protein concentrations do not fluctuate wildly between adjacent points in the cell cycle. The strength of this penalty, governed by a regularisation coefficient, was selected individually for each protein using leave-one-out cross-validation across three replicate experiments, balancing the competing risks of over-smoothing genuine dynamics and overfitting measurement noise.

The experimental foundation for the analysis came from time course proteomics experiments in which yeast cultures were synchronised in early G1 by centrifugal elutriation and then sampled every twenty minutes for nearly five hours. Protein concentrations were quantified for 3373 proteins using sixteen-plex tandem mass tag labelling and liquid chromatography-tandem mass spectrometry. After removing 101 proteins known to be asymmetrically distributed between mother and daughter cells, a violation of a key modelling assumption, the team applied their deconvolution to 3272 proteins. To separate genuine dynamics from artefacts, they filtered for solutions with both high signal-to-noise ratios and high peak-to-trough ratios, ultimately identifying 539 proteins with high-amplitude cell cycle-dependent behaviour, with concentration swings ranging from roughly 1.3-fold to as much as 32-fold.

Validation against established yeast biology lent strong credibility to the results. Hierarchical clustering of the 539 trajectories produced five groups with coherent functional enrichments: proteins involved in DNA replication and chromosome organisation peaked around S and G2 phases, respiratory and ATP synthesis proteins peaked near S/early-G2 consistent with known oxygen consumption patterns, amino acid biosynthesis enzymes peaked around START, and carbohydrate metabolism proteins peaked during phases requiring cell wall synthesis. Individual case studies were equally convincing. The deconvolved trajectory of Acc1, the rate-limiting enzyme of fatty acid synthesis, showed a sharp rise peaking in S/G2, matching prior Western blot data and coinciding with the known peak in lipid biosynthesis. Ergosterol synthesis enzymes peaked in G1, ribosomal proteins and ribosome biogenesis factors peaked around late G1, and proteins such as Hsl1, Mcd1 and Pds1 showed dynamics identical to decades of classical literature.

Beyond reconstructing concentrations, the team leveraged their trajectories to infer cell cycle-dependent transcription factor activity by combining the deconvolved data with documented regulatory relationships from the YEASTRACT database. The analysis recovered known regulators such as Ace2, whose activity peaked in daughter early G1, and Yox1, most active during S/G2, while also proposing intriguing new candidates. Activity of the stress-responsive factors Msn2 and Msn4 showed pronounced peaks in daughter early G1 and S/G2, a finding supported by independent time-lapse imaging of Msn2 nuclear localisation, and the Hap4 and Hap5 components of the Hap complex peaked around S/G2, potentially contributing to the metabolic switching between fermentation and respiration observed during the cycle. The researchers have released their complete dataset, computer code and uncertainty estimates as openly accessible resources, anticipating that the cell cycle-resolved proteome will become a key reference for future investigations into how metabolic oscillations emerge and how they exert control over the fundamental process of cell division.

Subject of Research: Cell cycle-dependent protein dynamics in budding yeast resolved by deconvolution of bulk proteomics

Article Title: Cell cycle-dependent protein dynamics in budding yeast resolved by deconvolution of bulk proteomics

Article References: Zylstra, A., Rovetta, M., Vedelaar, S. R., Bleischwitz, C., Fülleborn, J. A., van Oppen, Y., Markus, H. P., Korbeld, K. T., Calzati, E., Milias-Argeitis, A., Buczak, K., Schmidt, A., & Heinemann, M. (2026). Cell cycle-dependent protein dynamics in budding yeast resolved by deconvolution of bulk proteomics. Molecular Systems Biology. https://doi.org/10.1038/s44320-026-00241-6

Image Credits: AI Generated

DOI: 10.1038/s44320-026-00241-6

Keywords: budding yeast, cell cycle, proteomics, deconvolution, Saccharomyces cerevisiae, mass spectrometry, protein dynamics, metabolic oscillations, transcription factors, systems biology, Cell, cycle-dependent

Single-cell RNA sequencing has transformed biology by allowing researchers to measure the complete transcriptional activity of thousands of individual cells in a single experiment. This technology reveals extraordinary diversity within tissues, distinguishing cell types and even subtle cell states that bulk measurements would average away. However, scRNA-seq carries a fundamental limitation: during the dissociation process required to prepare samples, cells are physically separated from their tissue context. The resulting data describe what each cell is doing, but not where it was located. Spatial transcriptomics was developed precisely to solve this problem, preserving the spatial organization of cells within tissue sections while still capturing gene expression information. Yet even the most advanced spatial platforms typically profile fewer genes per cell or provide lower molecular resolution than dissociative single-cell methods, and most experiments capture only a single moment in time.

The new framework, developed by Mahdi Pursalim, Kaveh Kavousi, and Parisa Shooshtari, tackles the integration problem directly. The approach operates in two complementary stages. In the first stage, the method performs reference-based spatial deconvolution, a computational procedure that uses scRNA-seq data as a reference atlas to estimate which cell types are present in each spot or region of a spatial transcriptomics measurement. Because spatial platforms often measure mixtures of cells within each capture location, deconvolution is essential for untangling these mixed signals and recovering the underlying cellular composition. The second stage applies single-cell spatial mapping, transferring individual cells from the scRNA-seq dataset onto their most likely spatial positions within the tissue, guided by the gene expression similarities between the single-cell profiles and the spatial measurements.

What distinguishes this framework from previous integration efforts is its explicit treatment of time. The method is designed to work with time-series data—sets of scRNA-seq and spatial transcriptomics measurements collected at multiple, consecutive time points. By applying the deconvolution and mapping procedures independently at each time point and then comparing the results across consecutive stages, the framework quantifies how spatial cell-type distributions shift over the course of a biological process. These temporal changes are summarized using complementary abundance-weighted spatial descriptors, quantitative measures that capture both the proportion of each cell type and its spatial arrangement within the tissue. In parallel, the framework characterizes inferred cell-state dynamics by tracking changes in dominant cell-type assignments across consecutive time points, providing a computational estimate of how the state composition of the tissue evolves.

Recognizing that any computational method is only as trustworthy as its validation, the researchers subjected the spatial abundance component of their framework to rigorous simulation-based benchmarking. They generated synthetic spatial datasets with known, predefined cell-type compositions, using these simulated ground truths to quantitatively compare their approach against alternative spatial deconvolution and mapping methods. This evaluation strategy is critical in a field where real experimental data rarely come with complete knowledge of the true cellular makeup of every spatial location. The benchmarking results support the framework’s ability to recover simulated spatial cell-type abundance patterns, offering reassurance that the estimates it produces reflect genuine biological structure rather than computational artifacts.

To demonstrate the framework’s biological utility and generalizability, the team applied it to two systems at opposite ends of the biological spectrum: human glioblastoma organoids and embryonic chicken heart development. Glioblastoma organoids are laboratory-grown, three-dimensional structures derived from human cells that recapitulate key features of aggressive brain tumors, making them valuable models for studying cancer progression in a controlled setting. The framework revealed temporal patterns of spatial cell-type organization and inferred cell-state composition within these organoids that could not be detected through static analyses alone, offering a dynamic view of how the tumor-like cellular ecosystem reorganizes over time.

The second application, embryonic chicken heart development, represents one of the most dramatic examples of coordinated cellular behavior in biology. The heart forms through precisely choreographed movements of diverse cell populations, and disruptions to these processes underlie many congenital heart defects. When the researchers applied their framework to time-series data from developing chicken hearts, it captured temporal patterns in spatial cell-type redistribution and inferred cell-state dynamics that illuminate how the cellular architecture of this vital organ emerges during embryogenesis. Together, the two case studies—one disease-focused, one developmental—demonstrate that the framework is not tailored to a single biological context but can be applied broadly across systems where time-series single-cell and spatial data are available.

The authors are careful to frame their results with appropriate scientific caution, a nuance that matters for how the work should be interpreted. The framework captures what the researchers describe as spatiotemporal cell-type redistribution and inferred cell-state dynamics—computational measures derived from molecular data. These inferred quantities should be understood as descriptive computational measures rather than direct evidence of physical cell migration or lineage-validated state transitions. In other words, observing that the abundance of a cell type increases in a particular spatial region between two time points does not prove that cells physically traveled there; the change could also reflect differential proliferation, cell death, or changes in the sampling of the tissue. Similarly, apparent shifts in cell-state composition represent computational inferences rather than experimentally validated transitions of individual cells between states.

This caution reflects a broader truth about computational modeling of biological dynamics. Single-cell and spatial transcriptomics are destructive measurements—each sample is consumed in the process of profiling it—so reconstructing temporal behavior necessarily involves stitching together observations from different specimens assumed to represent the same underlying process. Despite this inherent limitation, the value of such models is considerable. They generate testable hypotheses about which cell populations move where, when specific state transitions occur, and how the spatial organization of tissues changes during development or disease. These hypotheses can then guide targeted follow-up experiments, such as live imaging or lineage tracing, that directly observe the behaviors the models predict.

The publication arrives at a moment when the experimental technologies underpinning the framework are advancing rapidly. Spatial transcriptomics platforms continue to increase in resolution, throughput, and affordability, and time-course studies combining scRNA-seq with spatial profiling are becoming increasingly common in developmental biology, cancer research, and regenerative medicine. What has lagged behind is the computational toolkit for integrating these data modalities across time in a principled, quantitative, and benchmarked way. By providing a broadly applicable framework with validated spatial abundance estimation, temporally resolved descriptors of cellular organization, and demonstrated performance across biologically distinct systems, the researchers offer the community a practical resource for systematic characterization of dynamic cellular behaviors. The work points toward a future in which the choreography of cells—their movements, positions, and transformations—can be reconstructed computationally from molecular snapshots, deepening understanding of how tissues build themselves, how tumors evolve, and how these processes might ultimately be modulated for therapeutic benefit.

The methodological choices behind the framework reflect practical constraints inherent to current spatial transcriptomics platforms. Reference-based deconvolution is necessary because many widely used spatial assays capture expression from groups of cells within each measured location, meaning the raw signal represents a mixture rather than a single cell. By anchoring this unmixing step to scRNA-seq references, the framework leverages the superior molecular resolution of dissociative profiling while retaining the positional information that spatial measurements provide. The complementary use of single-cell spatial mapping then adds a second layer of inference, assigning individual cells from the reference to plausible tissue positions based on expression similarity.

The benchmarking strategy deserves particular attention from readers evaluating the method. Simulation-based validation, in which synthetic spatial datasets are generated with known cell-type compositions serving as ground truth, allows quantitative comparison against alternative deconvolution and mapping approaches in a way that real tissue data cannot. Because no experimental measurement of a real tissue comes with a complete, verified inventory of every cell type at every location, simulations provide the only setting where estimation accuracy can be measured directly. The reported performance of the spatial abundance component against competing methods therefore offers a meaningful, if partial, assessment of reliability.

The two biological applications also illustrate the diversity of questions the framework can address. Glioblastoma organoids and the embryonic chicken heart differ not only in species and biological context but in the character of their cellular dynamics: one models the progressive reorganization of a tumor-like cellular ecosystem, while the other captures the choreographed emergence of organ architecture during embryogenesis. That a single computational pipeline can extract temporally resolved patterns of spatial cell-type organization and inferred cell-state composition from both systems suggests the approach is not dependent on tissue-specific assumptions, an important property for a tool intended for broad community use.

The authors’ framing of their results as descriptive computational measures rather than direct evidence of physical cell migration or lineage-validated state transitions is a notable feature of the work. Such interpretive restraint is uncommon and valuable in a field where language about cell movement can easily outpace what molecular snapshot data can actually demonstrate. The framework is published as open access under a Creative Commons license, and the article appeared as a citable, peer-reviewed accepted manuscript carrying a permanent DOI, with a final Version of Record to follow. The research was supported in part by the Children’s Health Research Institute and the Ontario Institute for Cancer Research.

Subject of Research: Integrative computational modeling of cell migration and tissue organization using time-series single-cell RNA sequencing and spatial transcriptomics data

Article Title: Systems-level modeling of cell migration using spatially-resolved single-cell data

Article References: Pursalim, M., Kavousi, K., & Shooshtari, P. (2026). Systems-level modeling of cell migration using spatially-resolved single-cell data. BMC Bioinformatics. https://doi.org/10.1186/s12859-026-06649-z

Image Credits: AI Generated

DOI: 10.1186/s12859-026-06649-z

Keywords: cell migration, spatial transcriptomics, single-cell RNA sequencing, spatiotemporal modeling, deconvolution, cell-state dynamics, glioblastoma organoids, chicken heart development, time-series analysis, computational biology, tissue organization, cell types