The review, conducted by researchers at Western New England University, Princeton University, Jahrom University of Medical Sciences, and Kurdistan University of Medical Sciences, followed PRISMA 2020 guidelines and screened 2,044 records from PubMed and Europe PMC covering literature from January 2010 through December 2025. After a two-stage automated screening process verified independently by two researchers, 123 studies met the inclusion criteria, each one combining an experimental microphysiological system with a computational model such as physiologically based pharmacokinetic modeling, quantitative systems pharmacology, in vitro-to-in vivo extrapolation, or machine learning. What emerged from the synthesis is a portrait of a field in explosive adolescence: publication output has been compounding at roughly 35% per year, and nearly 38% of all qualifying studies appeared in 2024 and 2025 alone.

To understand why this integration matters, it helps to grasp what each half brings to the table. Microphysiological systems are multicellular culture platforms engineered to recreate the physiological conditions of specific organs, complete with three-dimensional tissue architecture, mechanical forces such as stretch and fluid perfusion, and the cell types that make an organ function, including parenchymal cells, endothelial barriers, and immune cells. The field’s founding moment came in 2010, when Donald Ingber and colleagues at Harvard’s Wyss Institute built a lung-on-a-chip that reproduced the alveolar-capillary interface and demonstrated that cyclic breathing motions amplified inflammatory and toxic responses, something no static petri dish could ever capture. In the years since, chips have been built for virtually every major organ, and sophisticated multi-organ platforms now link human heart, liver, bone, and skin niches through recirculating vascular flow, generating pharmacokinetic profiles that begin to resemble those of an actual body.

But even the most elegant chip produces data that must be translated into predictions about intact humans, and that is where the computational half comes in. Physiologically based pharmacokinetic models represent the body as a network of compartments connected by blood flow, mechanistically predicting absorption, distribution, metabolism, and excretion from a drug’s physicochemical properties. Quantitative systems pharmacology models go further, folding disease biology and drug mechanism into mathematical frameworks that can forecast therapeutic outcomes. The digital twin concept binds these two worlds together through bidirectional information flow: the virtual model is initialized and calibrated with measurements from the physical system, and its predictions, in turn, guide which experiments to run next. In principle, this creates a continuously learning loop in which a chip’s human-relevant biology feeds a model that can simulate thousands of virtual drug trials before a single animal is touched.

The systematic review reveals exactly where this synergy has matured and where it remains embryonic. The liver dominates the landscape, appearing in 30.9% of studies, followed by vasculature at 19.5%, the gut at 18.7%, and immune components at 17.1%. This distribution is no accident: hepatic metabolism is the single most important determinant of drug clearance, and drug-induced liver injury remains one of the costliest reasons for late-stage clinical failures. Crucially, a gap analysis showed that only two organ-model pairings qualify as genuinely established, defined as more than five studies each: liver with physiologically based pharmacokinetic modeling, and liver with pharmacokinetic modeling. Countless combinations, particularly involving the kidney, lung, brain, skin, and pancreas, remain unexplored. Kidney, brain, and tumor models each appeared in just over 10% of studies, while lung, heart, and pancreas hovered near 5%, and skin registered a mere 1.6%.

The technical infrastructure underpinning the field is equally revealing. Among commercial platforms, Emulate led with 55.8% of citations, followed by Mimetas at 23.3% and CN Bio at 19.2%, yet custom and academic-built platforms still accounted for 74.2% of systems used, a sign that innovation remains distributed and that standardization is incomplete. On the software side, MATLAB was the most frequently reported computational environment, appearing in 20% of full-text analyzed studies, ahead of dedicated pharmacokinetic suites such as Simcyp, GastroPlus, and PK-Sim. Microfluidic designs dominated platform architecture at 76.7%, with polydimethylsiloxane as the fabrication material of choice in roughly half of studies, and 48.3% of platforms described explicit multi-organ configurations. Notably, adoption of multi-organ systems accelerated sharply after 2021, reflecting a growing recognition that organ-organ interactions, such as enterohepatic recirculation between gut and liver, are essential for accurate whole-body pharmacokinetic prediction.

Perhaps the most encouraging finding concerns validation, the make-or-break criterion for regulatory acceptance. A striking 95.8% of studies discussed comparisons with clinical or human data, and 77.5% reported quantitative accuracy metrics such as average fold error, the same standard regulators apply to physiologically based pharmacokinetic model qualification. The benchmark here is the Emulate Liver-Chip, which in a blinded assessment of 27 compounds recommended by the industry’s Innovation and Quality consortium achieved 87% sensitivity and 100% specificity for predicting drug-induced liver injury, dramatically outperforming the 50 to 60% concordance typical of animal toxicity studies. Animal models fail so often because species-specific differences in drug metabolism, immune responses, and disease mechanisms make rodent and non-rodent data unreliable guides to human outcomes, and methodological weaknesses like inadequate randomization and publication bias compound the problem.

The regulatory ground has shifted beneath all of this in ways that would have seemed unthinkable a few years ago. The FDA Modernization Act 2.0, signed in December 2022, eliminated the federal mandate for animal testing that had stood since 1938, explicitly defining cell assays, organ chips, microphysiological systems, and computer modeling as acceptable nonclinical tests. In September 2024, the first organ-on-a-chip, a Liver-Chip designed to predict drug-induced liver injury, was accepted into the FDA’s ISTAND pilot qualification program, and in April 2025 the agency announced a roadmap for phasing out animal testing in preclinical safety studies, beginning with monoclonal antibodies. Economic modeling based on the Liver-Chip’s validated performance suggests that routine adoption could save the pharmaceutical industry more than $3 billion annually by catching hepatotoxic drugs earlier, rescuing safe compounds wrongly flagged by insensitive assays, and accelerating go/no-go decisions.

Yet the review also punctures some hype with unsparing honesty. Despite the digital twin branding, most studies sit at the lower rungs of a maturity spectrum: the majority perform one-directional parameter transfer or one-time model calibration rather than the closed-loop, continuously updated coupling that defines a true twin. A manual review of the twelve machine-learning studies found that only one actually trained a predictive model on chip-derived data, a neural network predicting oxygen transport from vascularized chip morphology, while most used machine learning for image analysis or discussed it without applying it. And on the economics, the picture is nearly blank: while 92.5% of studies mentioned efficiency gains and 73.3% invoked animal reduction, only 10.8% cited dollar figures and not a single study performed a formal cost-effectiveness or cost-benefit analysis. The authors argue that liver-focused applications are now mature enough to serve as a template, and that the priority frontiers are kidney, lung, and brain integration, multi-organ systems paired with multi-compartment models, standardized parameter extraction, uncertainty quantification, shared benchmark datasets, and the health-economic evidence that will convince pharmaceutical decision-makers to abandon a century of dependence on animal testing.

Subject of Research: Integration of microphysiological systems with digital twin computational modeling for pharmaceutical development

Article Title: Integration of microphysiological systems with computational and digital twin modeling for pharmaceutical development: a systematic review

Article References: Soheili, M., Gilzad Kohan, B., Rastegar, L., Moradi, Y., & Gilzad Kohan, H. (2026). Integration of microphysiological systems with computational and digital twin modeling for pharmaceutical development: a systematic review. Journal of Pharmacy & Pharmaceutical Sciences, 29, Article 16454. https://doi.org/10.3389/jpps.2026.16454

Image Credits: AI Generated

DOI: 10.3389/jpps.2026.16454

Keywords: microphysiological systems, organ-on-chip, digital twin, PBPK modeling, drug development, pharmacokinetics, drug-induced liver injury, FDA Modernization Act 2.0, machine learning, IVIVE, preclinical testing, systematic review

The logic behind the approach, which the team calls the Macrophage Transducer Concept, draws on the natural biology of macrophages. These immune cells, derived from circulating monocytes, are among the first responders to emerging tumors. In early cancer, they detect factors released by cancer cells and migrate into tumor tissue, where interactions with malignant cells reshape their behavior. Tumor-associated macrophages then secrete molecules that influence inflammation, fibrosis, and the growth of new blood vessels, helping to build and regulate the tumor microenvironment. Crucially, macrophages are exquisitely sensitive to their surroundings and adjust their secretory output in response to diverse stimuli. The researchers hypothesized that this plasticity could be exploited: expose macrophages to a sample containing unknown cancer-derived factors, and the cells will translate that complex input into a reproducible profile of cytokines that can be measured with standard laboratory assays.

To test the idea, the team built a microfluidic device from polydimethylsiloxane, a transparent silicone widely used in lab-on-a-chip systems. The fabrication followed established soft-lithography techniques: a mold of the microchannel was patterned in SU-8 photoresist on a glass substrate using ultraviolet photolithography, liquid PDMS was poured over the mold and cured, and the resulting structure was plasma-bonded to glass. The finished channel is 500 micrometers wide and 200 micrometers deep, arranged in a serpentine pattern that packs 1.3 square centimeters of cell culture area into a compact footprint. A syringe pump drives fluid through the channel while a hot plate holds the device at body temperature, creating a controlled perfusion environment in which cells experience continuous, defined flow rather than sitting in a static pool of medium.

Before exposing the cells to cancer-derived factors, the researchers needed to understand how flow itself affects macrophage behavior. They cultured human monocyte-derived macrophages, prepared from blood donated by healthy volunteers, inside the device and perfused it with medium containing lipopolysaccharide, a bacterial endotoxin that strongly triggers tumor necrosis factor-alpha secretion. At flow rates of 1 and 2 microliters per minute, TNF-alpha output was high; at 5 microliters per minute it dropped significantly. Computational fluid dynamics using the finite volume method in ANSYS Fluent explained why. Simulating the velocity field ten micrometers above the channel floor—the height of the cells—revealed that the area of the channel experiencing flow velocities below 0.1 millimeters per second tracked closely with TNF-alpha secretion. Based on these results, the team selected 2 microliters per minute as the standard condition, balancing strong cellular responses with reasonable experimental throughput.

The perfusion format also unlocked something static cultures cannot provide: a time-resolved view of macrophage secretion. When LPS-containing medium was perfused through the device and effluent collected every four hours, TNF-alpha peaked sharply at four hours and then declined rapidly, while interleukin-6 rose to a maximum between four and eight hours and remained relatively stable afterward. In conventional well plates, by contrast, TNF-alpha did not peak until twelve hours and then plateaued, and IL-6 climbed gradually for roughly twenty hours. Because the two formats sample differently—perfusion captures what is secreted during each interval, while static culture accumulates everything up to the collection time—the absolute concentrations cannot be compared directly. But the contrast demonstrates that the microfluidic platform can resolve the onset, peak, and duration of secretory responses, dynamic information that endpoint measurements in well plates simply cannot capture.

With the platform calibrated, the team turned to the central question: do macrophages generate distinct, reproducible response profiles when exposed to different cancer-derived signals? They collected culture supernatants from eight human cancer cell lines representing colon, ovarian, esophageal, kidney, lung, breast, stomach, and liver cancers, and incubated them with macrophages for twenty-four hours. The answer was clearly yes. The types and amounts of factors secreted by the macrophages differed according to which cancer supernatant they had encountered. The colorectal line COLO205 drove elevated TNF-alpha and IL-6 secretion; the ovarian line RMG-1 provoked strong TNF-alpha and CCL2 responses; the kidney cancer line MAMIYA preferentially induced IL-6 and CCL2; and the gastric line MKN45 was characterized by increased vascular endothelial growth factor production. Four readout factors—TNF-alpha, IL-6, CCL2, and VEGF—were chosen to span inflammation, immune regulation, immune cell migration, and angiogenesis.

To visualize these multidimensional patterns, the researchers standardized the cytokine concentrations and generated heatmaps, then applied principal component analysis as an exploratory tool. The analysis suggested that macrophage-derived response profiles carry multidimensional information associated with differences among the cancer supernatants. Notably, when the pre-incubation supernatants were analyzed on their own, they showed one multivariate structure; after macrophage incubation, the spatial organization of samples in principal component space changed and dispersed, consistent with the idea that macrophages do not merely mirror the composition of the input but actively reshape and amplify biologically relevant information within it. The authors are careful to stress that the PCA was exploratory, performed on a limited donor dataset, and was not intended to demonstrate diagnostic classification performance. The goal was to establish whether complex biological information could be transformed into reproducible response profiles at all—not to identify the individual molecules responsible.

The team then confirmed that the concept works under perfusion. When supernatants from COLO205 and HepG2 liver cancer cells—chosen because they produced contrasting TNF-alpha responses in static culture—were flowed through the device, macrophages secreted TNF-alpha in response to COLO205 but not to HepG2, reproducing the static-culture trend. The device also proved economical: it reduced the number of cells required by 65 percent and the solution volume by roughly 10 percent compared with well plates, and TNF-alpha secretion per cell under LPS stimulation tended to be higher in the microfluidic format, measuring 33 percent higher under normalized conditions. The authors caution that the mechanisms behind this difference were not investigated, so any interpretation that the device enhances secretion remains speculative. They also note that integrating an on-chip biosensor, as other groups have done with fluorescence and electrochemical cytokine detection, could further shrink sample volumes and assay times, which are currently limited by the requirements of enzyme-linked immunosorbent assays.

The researchers are equally candid about the study’s limitations, which are substantial. Macrophages are heterogeneous, and although similar qualitative trends appeared across independent donors, the donor cohort was too small to systematically assess variability, and the differentiated cells were not phenotyped with markers such as CD80, CD86, CD163, or CD206. Only four secreted factors were measured, a small slice of the macrophage repertoire, and only two supernatants were tested in the device itself. The cancer-derived factors and signaling pathways driving the distinct responses remain unknown, PDMS surfaces may have adsorbed cytokines, and the culture supernatants used as model inputs differ enormously from real patient blood or serum. Disease specificity is also unresolved, since inflammatory and infectious conditions might produce overlapping macrophage responses, and no blinded validation was performed. Still, as a proof of concept, the work sketches a genuinely novel path for liquid biopsy: rather than fishing for known biomarkers in a sea of unknown signals, future diagnostics might hand complex clinical samples to a living immune sensor and read back the answer in the language of cytokines. Whether that translation proves reliable enough for the clinic will depend on the validation studies that must now follow.

Subject of Research: A microfluidic platform using macrophages as biological transducers to profile responses to cancer-cell-derived factors for potential cancer diagnostics

Article Title: Development of macrophage transducer-based cancer diagnostics: A proof-of-concept microfluidic platform for profiling macrophage responses to cancer-cell-derived factors

Article References: Ide, Y., Yamasaki, K., Ishikawa, S., Saito, Y., Fujiwara, Y., Nakanishi, Y., & Nakashima, Y. (2026). Development of macrophage transducer-based cancer diagnostics: A proof-of-concept microfluidic platform for profiling macrophage responses to cancer-cell-derived factors. Results in Engineering, 32, Article 113534. https://doi.org/10.1016/j.rineng.2026.113534

Image Credits: AI Generated

DOI: 10.1016/j.rineng.2026.113534

Keywords: macrophages, microfluidics, cancer diagnostics, liquid biopsy, cytokines, tumor microenvironment, biosensors, TNF-alpha, organ-on-chip, principal component analysis, tumor-associated macrophages, proof of concept

The platform itself is a feat of miniaturization. Fabricated from polydimethylsiloxane, the chip features a central culture chamber measuring just 2.5 by 1.3 millimeters, flanked by two lateral media channels. Into the chamber, the researchers injected a hydrogel of type I collagen at 6 milligrams per milliliter, seeded with human A549 lung adenocarcinoma cells at a density of 0.2 million cells per milliliter. Rather than pre-assembling spheroids and transferring them into the device, the team allowed the aggregates to form spontaneously within the matrix, guided entirely by microenvironmental conditions. This approach mimics a key aspect of early tumorigenesis, where isolated cells must organize themselves into multicellular structures before they can exploit the cooperative behaviors that make tumors dangerous.

Before perturbing the system, the researchers first characterized how glucose and oxygen behave inside the chip. By filling one lateral channel with high-glucose medium at 4.5 grams per liter and the other with glucose-free medium, they established a controlled gradient across the hydrogel. Glucose was detectable in the previously empty channel within 24 hours, and full equilibration between the two sides took about 48 hours, confirming that the collagen gel permits predictable diffusion. Oxygen, monitored with a fiber-optic dissolved oxygen probe at the channel inlets, drifted from roughly 10 parts per million down to a stable plateau near 7.5 parts per million over the first two days, a pattern that proved essentially identical whether cells were present or not. In other words, the chip’s intrinsic diffusion dynamics, rather than cellular consumption, dominate the oxygen profile at the device level, a detail the authors note as an important caveat when interpreting oxygen availability at the spheroid itself.

With the platform validated, the team launched a multifactorial experiment testing all eight combinations of glucose presence, glutamine presence, and normoxia or hypoxia. The outcome was unambiguous: glutamine is non-negotiable. When glutamine was removed, near-complete loss of cellular structures followed regardless of what else was available, and no spheroid-like formations could be detected by day 9. This aligns with a substantial body of literature showing that glutamine serves as a critical carbon and nitrogen source for biosynthesis, anaplerosis, and redox balance, and that blocking glutamine metabolism triggers apoptosis. Intriguingly, under hypoxia induced by sealing the chips with parafilm, some residual cellular growth persisted even without glutamine, hinting at a partial compensatory effect of low oxygen that the authors suggest may involve the exploitation of alternative metabolic routes, such as collagen-derived proline, though this proved insufficient to rescue full viability.

Glucose told a subtler and, in places, counterintuitive story. Spheroids deprived of glucose from the outset grew poorly, and withdrawing glucose at day 5 from cultures previously grown in low-glucose medium caused a marked arrest in growth. Yet, surprisingly, spheroids cultured continuously in low glucose at 1 gram per liter under normal oxygen levels grew faster than their high-glucose counterparts, a statistically significant difference. The authors attribute this to the known toxicity of elevated glucose concentrations, which can drive oxidative stress through reactive oxygen species production, glucose auto-oxidation, and the formation of advanced glycation end-products, alongside a hyperosmotic environment that disrupts proliferation and cytoskeletal organization. Under chemically induced hypoxia, however, the hierarchy flipped: spheroids in high glucose grew best, consistent with hypoxia-driven metabolic rewiring that upregulates glucose transporters and favors glycolytic flux.

The hypoxia signaling itself proved more nuanced than a simple oxygen meter. Using immunofluorescence for hypoxia-inducible factor 1-alpha, or HIF-1α, the researchers found that the only condition showing virtually no expression was normoxia with abundant glucose and glutamine, the metabolically comfortable state. But nutrient deprivation alone, even with normal oxygen, strongly induced HIF-1α, with the highest signals concentrated in the spheroid core and, under glutamine starvation, the strongest overall expression of all. Pimonidazole staining, which marks genuinely hypoxic tissue, revealed that these HIF-1α signals did not always track actual oxygen tension. The authors conclude that HIF-1α should be read as a marker of hypoxia-related signaling and metabolic adaptation rather than a direct proxy for local oxygen levels, a distinction with real consequences for how hypoxia is measured in cancer studies.

Perhaps the most visually compelling findings came from the structural analysis. Using confocal microscopy and 3D reconstructions from lattice light-sheet imaging, the team quantified actin organization, nuclear distribution, sphericity, and nearest-neighbor distances between nuclei. Well-fed spheroids in high glucose under normoxia were compact and beautifully layered, with a dense outer actin capsule, a proliferative periphery, and stable cellular protrusions extending into the matrix. Glucose withdrawal at day 5 dissolved those protrusions and spawned actin-rich knots in the interior, while chronic glucose absence produced spheroids that retained an outer shell but lost all internal organization, in some cases developing cavities at their cores. Remarkably, the external shape, captured by the sphericity metric, remained largely unchanged across glucose conditions, meaning that a tumor can look perfectly spherical from the outside while being hollowed out within.

Oxygen deprivation, by contrast, acted primarily on architecture rather than size. When hypoxia was induced at day 5, either chemically with ciclopirox olamine, a hypoxia mimetic that stabilizes HIF-1α, or physically by parafilm sealing, spheroids in high glucose showed no growth differences but suffered clear structural damage: loss of the peripheral actin capsule, reduced nuclear density at the rim, enlarged gaps between neighboring nuclei, and frequent internal cavities. Under glucose-free conditions, the combination proved devastating, with oxygen-depleted spheroids showing strong growth arrest and the most severe disorganization, including hollow, organoid-like structures. The authors interpret these cavities as most likely reflecting central necrosis or cell death, since live-dead staining showed spatial correspondence with ethidium homodimer-positive regions, though they caution that 2D imaging cannot definitively distinguish necrosis, apoptosis, or dysregulated autophagy as the underlying mechanism.

One finding carries particular therapeutic weight: when the researchers reintroduced depleted nutrients after prolonged deprivation, spheroids failed to resume normal growth, continuing to expand at the same suppressed rate for at least three days afterward. This suggests a form of metabolic memory, in which sustained nutrient stress leaves persistent adaptations, irreversible damage, or selected subpopulations that resist recovery. If such memory operates in real tumors, transient metabolic interventions could produce long-lasting suppression of tumor behavior well beyond the treatment window. The study has limitations the authors openly acknowledge, including reliance on a single cell line, the technical difficulty of measuring oxygen directly in the central chamber, and the low throughput inherent to sophisticated organ-on-chip systems. Yet the reproducibility of the trends across independent experiments and the platform’s demonstrated versatility with other cancer types in prior work suggest a robust foundation. As microfluidic tumor models mature, this chip offers something static cultures cannot: a dynamic, spatially resolved window into how tumors negotiate their metabolic environment, and a roadmap for exploiting the vulnerabilities that negotiation exposes.

Subject of Research: Effects of glucose, glutamine, and oxygen availability on the growth and architecture of lung adenocarcinoma spheroids in a microfluidic 3D collagen model

Article Title: Glucose and Oxygen Availability Shape Growth and Architecture of Tumor Spheroids in a Microfluidic 3D Model

Article References: Guerrero-López, P., & Garcia-Aznar, J. M. (2026). Glucose and Oxygen Availability Shape Growth and Architecture of Tumor Spheroids in a Microfluidic 3D Model. Annals of Biomedical Engineering. https://doi.org/10.1007/s10439-026-04377-2

Image Credits: AI Generated

DOI: 10.1007/s10439-026-04377-2

Keywords: tumor spheroids, microfluidics, cancer metabolism, glucose depletion, hypoxia, HIF-1α, glutamine, 3D cell culture, tumor microenvironment, lung adenocarcinoma, collagen hydrogel, organ-on-chip

The core problem the authors address is one of biological complexity. Glioblastoma is not a single disease but a mosaic of genetically and phenotypically distinct cell populations that coexist within the same tumor. Single-cell analyses have revealed remarkable cellular states, including stem-like populations and drug-tolerant persister cells that can survive initial treatment and later seed recurrence. Traditional two-dimensional cell cultures and even many animal models fail to capture this heterogeneity, which is one reason why therapies that look promising in the laboratory routinely fail in clinical trials. Rodent models, while valuable for studying tumor biology and the brain microenvironment, often do not faithfully reproduce the human tumor’s behavior or its response to drugs, and they are slow and expensive to run at the scale needed for individualized medicine.

Ex vivo platforms offer a way around these limitations. These systems take cells or tissue directly from a patient’s surgically removed tumor and grow them outside the body in conditions designed to mimic key aspects of the tumor’s native environment. The review highlights several complementary approaches at different stages of maturity. Patient-derived glioblastoma organoids, three-dimensional self-organizing cultures, have been shown to preserve both the inter-tumoral and intra-tumoral heterogeneity of the original cancers, and recent work has even used them as real-time avatars to test responses to clinical CAR-T cell therapies. Organotypic brain slice cultures maintain the tumor within a fragment of actual brain tissue, preserving interactions with neurons, glia, and immune cells. Microfluidic organ-on-chip systems go further still, incorporating flowing fluid, endothelial barriers that model the blood-brain barrier, and controlled gradients of oxygen and nutrients that shape drug penetration and tumor cell behavior.

Among the platforms the authors examine in detail is a microphysiological system developed to reproduce the phenotypical heterogeneity and drug resistance characteristic of glioblastoma, which can be rapidly biofabricated from patient-derived cells. The review also describes a high-throughput screening strategy that assesses drug cytotoxicity not only in the bulk population of tumor cells but also in the drug-tolerant persister cells that conventional screens tend to miss. This distinction matters because persister cells are increasingly recognized as a central mechanism of relapse across many cancers. A screen that kills the bulk tumor but leaves persisters untouched would paint a dangerously optimistic picture. By quantifying drug effects across both compartments, these assays aim to identify regimens capable of eliminating the full cellular spectrum of the disease. Notably, a randomized clinical trial using a related cancer stem cell assay to guide chemotherapy in recurrent glioblastoma reported improved survival for patients whose treatment was selected by the assay, providing early clinical evidence that functional testing can translate into patient benefit.

Artificial intelligence enters the picture as the analytical engine that can amplify the value of these platforms. Machine learning algorithms excel at finding patterns in large, multidimensional datasets, exactly the kind of data that ex vivo drug screens, genomic profiling, and imaging studies generate. The review outlines how AI approaches can enhance microphysiological platforms at multiple levels. In diagnostics, deep learning models applied to magnetic resonance imaging have reached the point where they can differentiate glioblastoma from lesions that mimic it, and uncertainty-aware ensembles of foundation models have been developed to make such classifications more robust. In treatment planning, AI-driven predictions of glioblastoma recurrence from metabolic and diffusion MRI are being used to define radiotherapy targets more accurately than conventional contrast-enhanced imaging allows. Machine learning and deep learning methods have also been applied systematically to prognosis, predicting survival and treatment response from combinations of imaging features and molecular data.

Perhaps the most forward-looking concept the review explores is the integration of drug screen outputs with AI-based modeling to create something approaching a medical digital twin for each patient. In this vision, the results of testing a patient’s tumor cells against a panel of drugs on a chip would be combined with the patient’s genomic, imaging, and clinical data, and machine learning models trained on large cohorts would then predict which regimen offers the best chance of efficacy with acceptable toxicity. Recent work on patient-specific prediction of glioblastoma growth using reduced-order modeling and neural networks demonstrates that computational models can already forecast individual tumor trajectories. A twin-tower architecture that fuses MRI and gene expression data to predict therapy response illustrates how multimodal AI can bridge the gap between laboratory measurements and clinical decisions. The authors suggest that combining efficacy, safety, and pharmacology results from drug screens with the output of extensive AI searches can yield insights that guide physician and patient decision-making and potentially improve prognosis.

The regulatory and scientific landscape appears to be shifting in favor of these human-based technologies. The United States Food and Drug Administration has announced a plan to phase out animal testing requirements for monoclonal antibodies and other drugs, and the National Institutes of Health has stated a priority to fund human-based research technologies. These policy changes could accelerate the adoption of organ-on-chip systems and organoid platforms in preclinical development, and by extension, their use in functional precision medicine. In hematologic cancers, functional precision medicine programs that test drugs against a patient’s living cells have already demonstrated clinical benefit in advanced aggressive disease and have identified exceptional responders, offering a template for what could be achieved in neuro-oncology.

The authors are careful to acknowledge the substantial challenges that remain. Standardization is a major hurdle: organoid cultures can vary between laboratories and even between batches, and microphysiological systems require sophisticated fabrication and handling that are not yet widely available. The turnaround time for growing and testing patient-derived cultures must fit within the narrow window between surgery and the start of adjuvant therapy, typically a matter of weeks. AI models trained on data from one population may not generalize to others, and uncertainty-aware methods are needed to flag when a model’s predictions should be trusted. There are also ethical considerations, including patient privacy, data governance, and the equitable access to technologies that will initially be expensive and concentrated in specialized centers. The review’s authors themselves hold patents and founding roles in companies developing these platforms, underscoring both the commercial momentum behind the field and the importance of independent validation.

What emerges from the review is a coherent argument that neither ex vivo platforms nor AI alone will be sufficient to change the trajectory of glioblastoma care, but that their integration could be transformative. The chip provides the biological ground truth, a living, patient-specific model in which candidate therapies can be tested under physiologically relevant conditions. The AI provides the pattern recognition and predictive power to interpret those results in context, drawing on vast datasets that no individual clinician could assimilate. Together, they promise a form of functional precision medicine in which treatment is selected not by tumor location and histology alone, but by demonstrated response of the patient’s own tumor cells, cross-validated by computational models. For a disease in which the standard regimen has remained essentially unchanged for two decades, and in which nearly every patient eventually relapses, the prospect of rapidly screening combinations that target both the bulk tumor and its drug-tolerant reservoirs represents a genuine shift in approach.

The path from review to routine clinical practice will require prospective trials demonstrating that AI-informed, assay-guided treatment selection improves outcomes beyond current standards, along with the infrastructure to run these platforms at scale. But the pieces are falling into place with unusual speed. Organoid biobanks are being established, organ-on-chip systems are becoming more reproducible and accessible, machine learning models for glioblastoma are advancing through systematic validation, and regulators are signaling openness to alternatives to animal testing. If the integration envisioned by the authors matures as hoped, a patient diagnosed with glioblastoma in the coming years may have their tumor grown on a chip within days of surgery, screened against dozens of drug regimens, and analyzed by algorithms that weigh efficacy, toxicity, and pharmacology to recommend the most promising option. For one of medicine’s most formidable cancers, that would mark the arrival of truly personalized treatment.

Subject of Research: Integration of ex vivo microphysiological platforms and artificial intelligence for personalized glioblastoma treatment selection

Article Title: Integrating ex vivo platforms with AI to guide glioblastoma treatment

Article References: Bardhan, M., Paul, S., Pun, S., Aryal, M. R., Claudio, P. P., Binder, J., Barrile, R., & Krummel, D. A. P. (2026). Integrating ex vivo platforms with AI to guide glioblastoma treatment. Journal of Neuro-Oncology, 179(2), Article 70. https://doi.org/10.1007/s11060-026-05708-y

Image Credits: AI Generated

DOI: 10.1007/s11060-026-05708-y

Keywords: glioblastoma, artificial intelligence, microphysiological systems, organoids, organ-on-chip, personalized medicine, functional precision medicine, drug screening, machine learning, blood-brain barrier, drug-tolerant persister cells, neuro-oncology

The appeal of lung-on-a-chip systems is straightforward. Conventional two-dimensional cell cultures strip away the organised tissue interfaces, fluid dynamics and mechanical forces that define the respiratory microenvironment, while animal models suffer from interspecies differences that often fail to predict human responses. By combining microfabrication and microfluidic control with pulmonary cell culture, chip platforms can reproduce epithelial-endothelial interfaces, air-liquid interface culture, vascular perfusion and even breathing-related strain, features first brought together in the landmark 2010 human lung-on-a-chip by Huh and colleagues at Harvard. Since then, the design space has exploded, encompassing immune-competent models, patient-derived cells, high-containment infection systems and inhalation exposure devices.

Yet the review’s central message is provocative: more complexity does not automatically mean better science. Membrane thickness, pore size, porosity, stiffness and extracellular matrix functionalisation are not passive structural details; they actively shape transport, cell attachment, mechanical deformation and signalling between compartments. Small-pore polyester membranes, for example, favour a compartmentalised epithelial barrier, while larger pores are needed if researchers want immune cells to migrate across the interface. Barrier fidelity and leukocyte recruitment therefore demand different, sometimes conflicting, interface properties, and choosing between them depends entirely on the question being asked.

Fluidic strategy introduces a similar set of trade-offs. Perfusion delivers nutrients, removes waste, transports drug compounds and mechanically stimulates endothelial cells, but volumetric flow rate alone does not define the cellular environment, because wall shear stress also depends on channel geometry and fluid viscosity. Pump-driven systems offer precise control over flow profiles but increase dead volume, bubble risk and operational burden. Pump-free, gravity-driven rocking platforms simplify operation and reduce contamination risk in high-containment viral infection studies, at the cost of less precise control over instantaneous flow. The authors argue that studies should routinely report the flow-generation method, channel dimensions and calculated or measured shear stress so that perfusion-mediated transport can be distinguished from endothelial mechanostimulation.

Mechanical actuation, the defining capability of breathing chips, receives particularly careful treatment. The review emphasises that breathing motion is not a single condition but a family of variables spanning strain magnitude, frequency, waveform and actuation mechanism. Recent studies show these parameters matter enormously: cyclic stretch combined with airflow can accelerate mucociliary maturation of airway epithelium, and dynamic strain can modify how cells respond to inhaled toxicants. In one striking example, an orthotopic lung cancer chip demonstrated that breathing-related deformation altered tumour proliferation, invasion, dormancy and even responsiveness to tyrosine kinase inhibitors, implicating mechanical cues in epidermal growth factor receptor and MET signalling. Therapeutic response, in other words, cannot always be attributed to cellular genotype and drug exposure alone when the mechanical microenvironment itself modifies phenotype.

The same principle applies to exposure delivery and cellular composition. A next-generation breathing chip integrated with whole-cigarette-smoke exposure reported an approximately sixty percent reduction in barrier resistance and a 4.5-fold increase in interleukin-8 expression, with dynamic strain accelerating barrier disruption relative to static conditions, results that submerged chemical extracts could never reproduce. Cellular complexity, meanwhile, should be justified functionally rather than maximised blindly. An immune-competent lung-on-a-chip modelling severe influenza found that adding tissue-resident macrophages generated cytokine profiles more closely matching patient bronchoalveolar lavage data, but reproducing the broader inflammatory storm of severe infection required resident and circulating immune populations, perfusable microvasculature and a five-micrometre-pore interface that permitted immune migration. Complexity earned its place by enabling a predefined function.

Manufacturing and quality control emerge as underappreciated determinants of credibility. Small variations in channel dimensions, membrane position, bonding or actuator geometry can alter shear stress, strain transfer and barrier behaviour, and such technical variation can easily masquerade as biological variability. The review calls for predefined tolerances on critical device attributes, functional quality-control criteria including leak testing and sensor calibration, batch traceability, and documented fabrication parameters. It also cautions that switching fabrication methods, from soft lithography in polydimethylsiloxane to thermoplastic replication for scalable production, can change gas permeability, surface chemistry and compound adsorption, and should therefore trigger re-verification of critical engineering and biological endpoints rather than being treated as a neutral process change.

To organise the path from bench to regulatory relevance, the authors propose a five-stage framework anchored to a clearly defined context of use. The first stage, engineering verification, confirms that the device reproducibly generates its intended physical environment, from verified strain fields to delivered smoke exposure. The second, biological qualification, demonstrates that this environment supports the specific biological functions the application requires, such as barrier integrity, mucociliary activity or immune-cell migration. The third stage, disease or pharmacological validation, requires reproducible detection of the target response against appropriate comparators, as illustrated by a human alveolus chip that reproduced radiation-induced DNA damage, inflammation and barrier disruption and then evaluated prednisolone and lovastatin as candidate countermeasures.

The final two stages demand the most stringency. Human concordance requires direct comparison of chip outputs with patient-derived or clinical reference data, a standard met in part by the influenza chip’s cytokine benchmarking against patient samples, but the authors warn that such evidence should not be stretched into broader claims of clinical prediction. Predictive performance, needed when a context of use involves therapeutic or toxicological decision-making, requires predefined metrics such as sensitivity, specificity or response discrimination against characterised reference datasets. Deployment and regulatory readiness then asks whether qualified performance survives manufacturing scale-up, automation and transfer between independent laboratories. The framework deliberately distinguishes physiological resemblance from demonstrated concordance, and concordance from clinically anchored prediction, so that each claim carries only the weight its evidence supports.

The review’s conclusion is a call for discipline over dazzle. Prioritising validated, fit-for-purpose performance over maximal complexity, the authors contend, offers a more credible route toward reliable respiratory research tools, drug development platforms and eventual regulatory decision support. Standardisation, they argue, should target terminology, reporting of critical parameters, reference controls and inter-laboratory reproducibility rather than forcing identical device architectures, since legitimate applications genuinely require different designs. As regulators, including the United States Food and Drug Administration in recent draft guidance, increasingly frame new approach methodologies around context of use, the framework provides developers with a practical checklist: define the question and acceptance criteria before experimentation, report engineering parameters with tolerances, benchmark against prespecified comparators, incorporate human datasets when predictive claims are intended, and prove that performance holds across batches, operators and laboratories. If the field follows that path, the miniature lungs growing on chips today may earn the trust needed to influence tomorrow’s drug decisions.

Subject of Research: Engineering and translational validation of lung-on-a-chip microphysiological devices for respiratory disease modelling and drug testing

Article Title: Engineering Lung-on-a-chip microdevices for respiratory disease modelling and drug testing: A fit-for-purpose framework for design and translational validation

Article References: Cheong, S. Y., Liew, T. I. Z., Wong, C. K., Teng, X. X., Cha, X. Y., Ng, N. C.-S., Wong, R. S.-Y., & Goh, B. H. (2026). Engineering Lung-on-a-chip microdevices for respiratory disease modelling and drug testing: A fit-for-purpose framework for design and translational validation. Biomedical Microdevices, 28(3), Article 62. https://doi.org/10.1007/s10544-026-00849-3

Image Credits: AI Generated

DOI: 10.1007/s10544-026-00849-3

Keywords: lung-on-a-chip, microfluidics, organ-on-chip, respiratory disease modelling, drug testing, microphysiological systems, translational validation, mechanical strain, inhalation toxicology, quality control, regulatory science, context of use