Comparison of experimental tumor growth curves with model predictions to define baseline values of model parameters.

Accurate respiratory monitoring in laboratory mice is crucial to ensure animal safety, optimize the anesthetic dosing, enable the timely detection of drug-related adverse events, and assess animal physiological conditions during experiments. Several methods have been proposed for monitoring respiratory rate (RR) in laboratory mice, differing in terms of invasiveness, measuring principle, and accuracy. Among non-invasive techniques, small sized and soft wearable sensors represent a promising solution for RR monitoring in anesthetized mice, thanks to their ease of use, and suitability for long-term and continuous monitoring. However, their potential for RR monitoring in mice remains largely unexplored.The present study proposed a soft wearable sensor (WS) based on fiber Bragg grating (FBG) technology for RR monitoring in anesthetized mice. FBG sensors, which are already well established in human respiratory monitoring, offer high sensitivity to small deformations, immunity to electromagnetic interference, compact size, and multiplexing capabilities. These features make them particularly suitable for application in murine models characterized by high respiratory frequencies and low-amplitude chest wall motions. Here, we first carried out the design, fabrication, and metrological characterization of the proposed WS. Then, a feasibility assessment was conducted to evaluate the WS capability to detect breathing-induced deformations and extract RR in anesthetized mice. The results revealed the potential of the proposed FBG-based WS to non-invasively monitor RR in preclinical murine models.

BACKGROUND:Antibody-drug conjugates (ADCs) and bispecific antibodies represent a rapidly advancing frontier in oncology, yet the abnormal tumor microenvironment (TME) hinders their delivery and reduces efficacy. Emerging immunomodulatory ADCs (IM-ADCs) demand mechanistic mathematical models that couple drug transport with immune dynamics. METHODS:Here, we present a mechanistic framework for the delivery of HE-S2 ADC, an anti-programmed cell death ligand 1 (PD-L1) antibody bearing the bifunctional immunomodulator D18. Our model integrates cancer-immune cells interactions, TME properties, such as dysfunctional vessels, elevated interstitial fluid pressure, tissue hydraulic conductivity, and vascular permeability, spatiotemporal distributions across growing tumor and adjacent host tissue, convective-diffusive transport, ADCs binding and internalization kinetics and tumor-draining lymph node biology governing antigen presentation and the generation of effector CD8+ T cells. Parameters were calibrated simultaneously with the murine MC38 and B16 tumor growth data and effector CD8+T cell data following treatment with D18, anti-PD-L1, and ADC. RESULTS:Our mechanistic spatiotemporal model captures the superior antitumor efficacy of the HE-S2 ADC relative to its individual components and provides mechanistic predictions for unmeasured variables, such as spatiotemporal dynamics of drug/immune-cell distributions. It explains reduced intratumoral D18 exposure via rapid clearance, while antibody/ADC achieves higher tumor retention through leaky tumor vasculature. The model suggests a reinforcing loop in which improved ADC exposure enhances CD8+T cell infiltration, driving tumor shrinkage that lowers fluid pressure and improves drug delivery. Parametric analyses findings support TME normalization strategies that increase functional vessel density prior to ADC administration; however, such approaches should preserve sufficient vascular permeability by maintaining vessel pore radius >~40 nm, ensuring pores remain large enough for ADC extravasation and effective intratumoral delivery. CONCLUSION:The proposed mechanistic model successfully captures how TME properties regulate the delivery and efficacy of IM-ADCs while suggesting TME normalization as a potential strategy to improve treatment outcomes.

Precise prediction of cancer therapy response remains challenging because conventional biomarkers capture molecular features but overlook the physical state of tumors. We developed a multimodal deep-learning framework integrating ultrasound shear wave elastography (SWE) images with quantitative stiffness measurements (elastic modulus, kPa) to predict treatment outcomes in preclinical murine tumors. Each image–stiffness pair is tokenized within a transformer that learns interactions between local elastographic texture and global rigidity. A lightweight convolutional encoder extracts image features, the modulus is embedded as a numeric token, and self-attention fuses both modalities for classification. Trained on 1,578 baseline SWE images from five syngeneic tumor models, the model classified tumors as responders, stable, or non-responders. Across five random seeds, it achieved 92.4% ± 1.3% accuracy, macro-F1 0.92, and ROC-AUC 0.99 on a held-out test set, with well-calibrated probabilities. Matched-split ablations—image-only, stiffness-token-removed, stiffness-shuffled, late-fusion, and stiffness-only—showed that performance reflected genuine cross-modal learning rather than scalar stiffness alone; shuffling image–stiffness pairings significantly reduced accuracy (all corrected p < 0.01). Leave-one-tumor-model-out analysis demonstrated generalization to unseen tumor types, with 95.5% ± 1.5% accuracy (range 93.9–97.9%) across five held-out models. Lower baseline stiffness correlated with better response, supporting the hypothesis that mechanically normalized tumors respond more effectively. These preclinical proof-of-concept findings establish tumor mechanics as candidate predictive biomarkers and transformer-based multimodal learning as a scalable approach to biomechanically informed response prediction, while requiring validation in human cohorts.

Glioblastoma multiforme (GBM) remains a devastating disease with a poor prognosis. GBM progression generates mechanical forces that compromise blood vessels' functionality, impair perfusion, and create hypoxic regions, which in turn reduce radiosensitivity. While the link between hypoxia and radiosensitivity is known, a quantitative methodology to predict this effect is lacking. Here, we developed a patient-specific mechanistic mathematical model of radiotherapy that integrates Magnetic Resonance Elastography (MRE) imaging. The model can translate MRE-derived stiffness into spatial maps of mechanical stress and simulate subsequent events, such as compression of vessels and impaired intratumoral oxygen distribution which impacts radiosensitivity. Our simulations show that biomechanical properties of both tumor and host tissue control patterns of radiosensitivity. Notably, heterogeneous distribution of tumor elastic properties as well as the presence of regions of higher host-tissue stiffness adjacent to the tumor boundary, generate increased and heterogenous mechanical stresses within the tumor. The results are compressed vessels with hypo-perfused and hypoxic tumor tissue, and consequently compromised radiotherapy efficacy. Results show the importance of tumor microenvironment (TME) parameters and suggest that strategies to normalize the TME could improve treatment outcomes and help stratify treatments. Taking together, our work establishes a quantitative pipeline linking MRE biomechanics to oxygen-modulated radiosensitivity.

Pancreatic ductal adenocarcinoma (PDAC) develops within a biomechanically abnormal tumor microenvironment, characterized by a dense stroma and elevated compressive forces. While extracellular matrix stiffness has been extensively studied, the impact of compressive forces on immune regulation and tumor–immune interactions remains poorly understood. We integrated two complementary bioengineered compression models, a 2D transmembrane pressure device and confined 3D spheroids, with bulk transcriptomic and Liquid Chromatography–Mass Spectrometry (LC-MS)–based exometabolomic profiling to examine how mechanical compression shapes macrophage behaviour and tumor–immune crosstalk. Controlled compressive stress (0–8 mmHg) was applied to macrophages, tumor cells, and tumor–macrophage cocultures, followed by pathway analysis, functional assays, and multi-omic integration. Mechanical compression activated conserved mechanotransduction pathways in macrophages, including PI3K/Akt and MAPK/SAPK signaling, and induced transcriptional programs associated with inflammatory activation consistent with an M1-like macrophage phenotype and cytoskeletal remodelling. In parallel, compressed tumor cells adopted an immunomodulatory state marked by increased expression of immunosuppressive cytokines and macrophage checkpoint signals. When tumor cells and macrophages were simultaneously exposed to compression, functional assays revealed a shift of macrophages toward immunosuppressive phenotypes. Bulk RNA sequencing identified cell-type–specific transcriptional responses converging on metabolic pathways, while LC-MS exometabolomics revealed compression-dependent enrichment of extracellular nucleotide metabolites in tumor–macrophage cocultures. These findings identify compressive stress as a critical regulator of immune suppression and tumor–immune metabolic coupling in PDAC, highlighting mechanical forces as important drivers of immune dysfunction in mechanically constrained tumors.

Schematic of the mathematical model of tumor-immune system interactions, highlighting key immune components, their activation and cytolytic functions, and modulation by PD-1/PD-L1 signaling. The model also incorporates cytokine effects and the influence of immunosuppressive cells like M2 macrophages and Tregs.

IntroductionLocal radiotherapy rarely triggers regression of distant, non-irradiated tumors (the “abscopal” effect), but this outcome is unpredictable because it depends on interacting processes, such as antigen release, antigen presentation, T-cell priming and trafficking, and lymphoid health. To study these interactions quantitatively and identify dominant mechanisms that control off-target tumor responses, we built an integrated physiologically based pharmacokinetic - quantitative systems pharmacology (PBPK-QSP) model.MethodsThe PBPK-QSP model tracks immune (dendritic cells, M1/M2 macrophages, Tregs, naïve and effector CD8+ T cells, Antigen Presenting Cells) and tumor cell populations across body compartments that include a local (irradiated) and distant tumor, along with major organs, as well as the blood and lymph circulations. Radiotherapy is considered to have local effects on direct tumor cell killing, and indirectly by releasing tumor-associated antigens that induce immune cell priming, which drives the abscopal effect. Furthermore, we tested how lymph-node irradiation can affect immune cell priming. The model was calibrated against pertinent preclinical data, and sensitivity and correlation analyses were performed to investigate the mechanisms of off-target tumor response. ResultsFour mechanisms dominate outcome variability: antigen capture/processing by phagocytes (i.e., dendritic cells and macrophages), clearance of dead-cell debris and antigens, and naïve T-cell regenerative capacity in lymph nodes. Phagocytic and clearance rates have context-dependent effects, too fast shortens the antigen-priming window, and too slow results in less overall antigen-priming. Lymph-node irradiation shifts the dependence of immune response to T-cell recovery, which becomes the dominant mechanism. The model also highlights that impaired tumor vascular permeability can constrain effector infiltration and mute intratumoral CD8+ T cells differences between the local and distant tumor despite systemic activation.DiscussionThe PBPK-QSP model identifies specific, actionable mechanisms controlling abscopal responses and suggests three complementary strategies to increase the chance of abscopal responses: i) optimize radiotherapy dose/fractionation to maximize immunogenic antigen release while sparing lymphoid tissue when possible, ii) combine radiotherapy with interventions that prolong productive antigen presentation and modulate debris clearance, and iii) protect/restore lymphoid regenerative capacity.

Supplementary raw data of the results of the paper

Global sensitivity analysis. The coefficients of significant predictors (p-value<0.05 ANOVA) on the tumor volume before treatment in equation (3) of the main paper.

Comparison of the components incorporated in our modeling framework with those incorporated in previously published pertinent models.

Atomic Force Microscopy (AFM) is a key method for nanomechanical characterization of cells and tissues, with AFM-derived fingerprints proposed as biomarkers for cancer diagnosis and treatment monitoring. These signatures typically include a higher elasticity peak (HEP), reflecting extracellular matrix stiffening due to collagen overproduction, and a lower elasticity peak (LEP), indicative of cancer cell softening. Despite their potential, AFM elasticity spectra are often assessed qualitatively, and a standardized mathematical framework for quantitative analysis is lacking. Here, we provide a rigorous mathematical characterization of Young’s modulus distribution in normal and cancerous tissues, aiming to improve cancer diagnosis and treatment monitoring. Previously published AFM elasticity spectra from murine tumors were employed and analyzed to evaluate tumor nanomechanical changes at different time points, 14, 21, and 28 days after cell implantation, including both untreated controls and a tranilast-treated group, with tranilast being a drug known to reduce collagen levels. The weighted skew-normal distribution was employed to model AFM data due to its ability to capture the two-peak structure of cancerous tissue, reflecting a mixture of soft cancer cells and stiffer components. We hypothesized that tranilast treatment would progressively shift the HEP to lower values. Model accuracy was confirmed by high R² values and low Cramér–von Mises (CvM) criteria. Results revealed a transition from a two-peak distribution in controls (HEP and LEP) to peak convergence in tranilast-treated tissue at 28 days. We conclude that the weighted skew-normal distribution offers a robust method for quantifying tumor nanomechanics, which is related to therapeutic outcomes.

IntroductionTumor heterogeneity poses a significant challenge for predicting responses to cancer therapy, highlighting the need for the development of biomarkers to guide personalized treatment. Contrast-enhanced ultrasound (CEUS) imaging is an established method to assess tumor perfusion, which directly affects drug delivery and therapeutic efficacy, as poorly perfused tumors often limit the penetration of chemo- and immunotherapeutics.MethodsWe developed a deep learning framework using CEUS imaging to predict the response of tumors to chemo-immunotherapy in murine models of breast cancer, fibrosarcoma, and melanoma. A convolutional neural network (CEUS-CNN) was trained on a dataset of 587 pre-treatment CEUS images to classify tumors as responsive, stable, or non-responsive based on RECIST version 1.1 (Response Evaluation Criteria in Solid Tumors) criteria (175 responsive cases, 136 stable, and 276 non-responsive). Additionally, synthetic data were created for the responsive and stable classes to address class disparity.ResultsOur framework attained an overall test accuracy of 0.877 (0.941 for responsive, 0.615 for stable, 0.963 for non-responsive) using only real data. The addition of synthetic data led to improved model performance, with a notable impact on the previously underperforming stable class. Our strategy enhanced the predictive capability of our model, raising the average test accuracy to 0.930 (1.000 for responsive, 0.769 for stable, 0.963 for non-responsive).ConclusionThese findings support CEUS imaging as a possible imaging biomarker of response to cancer therapy and further indicate that the incorporation of synthetic data can enhance model effectiveness, particularly for underrepresented classes. Together, they highlight the potential value of integrating AI with CEUS for personalized cancer treatment strategies.

ABSTRACT Transgelin is an actin‐binding protein that promotes cancer progression via activation of cancer‐associated fibroblasts and has been identified as a prognostic marker. However, its distribution and functional role in colon cancer remain unclear. In this study, we aimed to elucidate the mechanistic role of transgelin in colon cancer progression by focusing on its functional impact in cancer‐associated fibroblasts. Tissue microarrays from 359 human colon cancer tissues were investigated to elucidate the clinical importance of transgelin expression in cancer stroma. We focused on transgelin in fibroblasts and investigated its functional role in stromal activation using in vitro knockdown experiments and in vivo co‐transplantation models. Primary cultures of human colon fibroblasts were evaluated for their biological function. Our data showed that transgelin expression is predominant in activated cancer‐associated fibroblasts in colon cancer tissues. Stimulation by cancer‐cell‐conditioned medium (CM) significantly upregulated transgelin, ACTA2, COL1A1, and TNC expression in colonic fibroblasts. Additionally, transgelin knockdown (KD) in fibroblasts did not influence the upregulation except for transgelin itself. Transgelin KD in fibroblasts did not result in drastic alterations in gene expression profiles. Transgelin KD suppressed collagen gel contractility. Furthermore, co‐transplantation experiments of cancer cells and colonic fibroblasts into immunodeficient mice revealed that transgelin KD inhibited tumor growth in fibroblasts. In conclusion, stromal transgelin expression in colon cancer strongly correlated with distant metastasis and served as a prognostic factor for colon cancer. Mechanistically, transgelin in cancer‐associated fibroblasts promotes tumor growth by regulating stromal contractility, suggesting transgelin as a potential therapeutic target.

Accurate temperature (T) monitoring is essential during medical procedures based on energy delivery to biological tissues to ensure both treatment safety and efficacy. Fiber Bragg grating sensors (FBGs) represent a well-established optical technology for T monitoring, offering high sensitivity, fast response, multiplexing capability, and immunity to electromagnetic interference, which make them particularly suitable for complex and challenging medical scenarios. Although FBG-based systems have been widely proposed for T measurements during thermal therapies (e.g., ablation procedures), their in vivo validation remains limited.In this work, we investigated the use of the FBG technology for in vivo T monitoring during focused ultrasound (FUS), a non-invasive strategy to concentrate acoustic energy within defined tissue volumes and induce controlled biological effects. The proposed sensing approach was assessed in murine breast tumor models, enabling real-time and localized T measurements directly within the tumor during ultrasound exposure. Results revealed a limited and gradual T increase, consistent with the predominantly non-thermal nature of the FUS approach, demonstrating the FBG capability to detect even modest T variations in vivo during ultrasound exposure.The proposed monitoring strategy has the potential to be readily extended to other ultrasound-based therapeutic procedures, such as high-intensity focused ultrasound (HIFU) ablation procedures, where substantially higher T increments are expected, and accurate in vivo T monitoring is critical for the treatment efficacy and the safety of patients.

Precision oncology frameworks primarily rely on tumor-intrinsic molecular features to guide therapeutic decision-making, yet the contribution of mechanical forces within the tumor microenvironment remains poorly defined. Here, we investigate how compressive stress, a fundamental mechanical force arising during tumor growth, reshapes tumor cell states and modulates clinically relevant therapeutic pathways. Using a controlled compression system, we comprehensively profiled transcriptional responses in pancreatic cancer cells and integrated these with datasets from mechanically compressed breast and liver cancer models. Compression induced extensive, tumor type–specific transcriptional reprogramming, characterized by activation of stress-adaptive pathways and suppression of proliferative programs. Gene-level overlap was limited between tumor types, though partial convergence emerged at the pathway level. Compression-associated transcriptional signatures were independent of pre-existing genomic alterations, yet detectable in patient tumors across TCGA cohorts, underscoring their clinical relevance. Higher compression signature scores were associated with reduced survival in pancreatic and liver cancers, but not breast cancer, revealing tumor type–specific prognostic associations. Integration with pharmacogenomic databases showed that compression-responsive genes intersect with multiple drug-target networks in pancreatic cancer, including inflammatory, angiogenic, and kinase signaling pathways. Complementary metabolomic profiling identified coordinated alterations in purine metabolism, sphingolipid signaling, and redox pathways, further linking mechanical stress to drug-relevant cellular processes. Together, these findings identify tumor compression as a regulator of molecular states that interface with therapeutic targets and patient outcomes. Incorporating compression-associated molecular signatures into precision oncology frameworks may improve the identification of therapeutic vulnerabilities and prediction of treatment response in solid tumors.

Local sensitivity analysis. The logarithm of total variance for each parameter and treatment.

Model validation with clinical dataset.