The team previously developed “Panregza,” a test that combines peripheral whole-blood gene expression patterns with the serum tumor marker CA19-9. While Panregza has shown utility in later-stage disease, its performance in stage 0–Ⅰ cancers—where tumor burden is minimal—had not been established.

In the current pilot case–control study, the researchers analyzed whole-blood mRNA expression using a panel of 56 gene probes. They evaluated stage 0–Ⅰ pancreatic cancer samples from 10 patients (about 4% of a larger cohort) and compared them with 104 healthy individuals. Diagnostic performance was assessed for (1) gene expression alone, (2) CA19-9 alone, and (3) the combined Panregza system.

The blood gene expression method identified 9 of 10 early-stage cases, corresponding to 90% sensitivity. By contrast, CA19-9 detected only 1 of 10 cases, or 10% sensitivity, highlighting the limitation of relying on tumor marker levels for early disease.

When CA19-9 was combined with the gene expression readout, the Panregza system achieved 60% sensitivity and 93.3% specificity. Together, these results suggest that peripheral whole-blood transcriptional signatures carry clinically meaningful information even when CA19-9 is normal.

Importantly, the findings support a biological model in which pancreatic cancer-associated signaling alters gene expression in immune and other blood cell populations. Because these changes can occur before substantial tumor growth, they may enable detection that is not dependent on tumor volume.

The study also emphasizes the clinical significance of early diagnosis. At Kanazawa University Hospital’s Innovative Research and Development Center for Pancreatic Cancer, reported five-year survival rates are 100% for stage 0 and 74.4% for stage Ⅰ, reflecting the impact of catching disease early.

Overall, the work provides viral-science-news momentum for whole-blood mRNA diagnostics in pancreatic cancer and strengthens the case for further validation toward scalable screening.

Subject of Research: Whole-blood mRNA expression diagnostic system for early-stage pancreatic ductal adenocarcinoma (Panregza)

Article Title: Pilot validation of a whole-blood mRNA expression-based diagnostic system for early-stage pancreatic ductal adenocarcinoma: a single-center case–control diagnostic accuracy study

News Publication Date:

Web References: https://doi.org/10.1038/s41598-026-58684-8

References: 10.1038/s41598-026-58684-8

Image Credits: © Kanazawa University

Keywords: pancreatic cancer; early detection; whole-blood mRNA; gene expression; CA19-9; diagnostic accuracy; biomarker; Panregza; sensitivity; specificity; immune-associated transcriptional signatures

Pancreatic cancer has long been notorious for its elusive early presentation and poor prognosis, with the majority of patients diagnosed only once the disease has metastasized. This late detection contributes to dismal survival statistics, with five-year survival rates languishing below 15%. The urgency of developing earlier diagnostic tools is underscored by projections indicating pancreatic cancer will become the second leading cause of cancer-related deaths in the United States by 2030. The Mayo Clinic’s REDMOD AI system directly addresses this unmet need by leveraging advanced computational techniques, including radiomics, which quantitatively analyze tissue texture, heterogeneity, and morphology from cross-sectional imaging.

The innovation embedded in REDMOD lies in its ability to extract and interpret hundreds of quantitative imaging features that capture the biological microenvironmental alterations preceding macroscopic tumor formation. Essentially, this approach translates the scanning data into high-dimensional datasets that delineate subtle phenotypic shifts in pancreatic tissue. Importantly, the AI model was trained and validated on nearly 2,000 CT scans sourced from multiple institutions, encompassing various imaging hardware and acquisition protocols. This rigorous multi-institutional validation ensures that the AI’s diagnostic precision is not limited by scanner variability or institutional bias, highlighting its potential for broad clinical deployment.

Furthermore, the Mayo Clinic team emphasized REDMOD’s longitudinal stability — the model generates consistent diagnostic outputs across sequential scans taken months apart from the same patient. This feature is critical for monitoring high-risk individuals over time and could facilitate ongoing surveillance strategies to detect the earliest signs of neoplastic transformation. Of particular note, REDMOD automatically processes routine abdominal CT images obtained for disparate clinical indications, obviating the need for specialized imaging or labor-intensive manual annotations. This design inherently supports integration into existing clinical workflows, enabling opportunistic screening and risk stratification.

The model’s performance was starkly superior to that of expert radiologists when analyzing scans originally interpreted as normal. REDMOD identified 73% of pancreatic cancers that were retrospectively known to be present up to 16 months before clinical diagnosis, effectively doubling the detection rate achievable through conventional human review alone. The detection advantage expanded further at earlier time points; more than two years prior to diagnosis, the AI model uncovered nearly three times as many hidden cancers as specialists. These compelling metrics underscore the transformative potential of AI-enhanced radiomics in addressing the longstanding challenge of occult pancreatic cancer.

REDMOD’s application could carry particular significance for high-risk populations, such as individuals with new-onset diabetes or a familial predisposition to pancreatic cancer. These groups often undergo abdominal imaging for various indications, creating an opportunity for “second-look” AI analysis to identify early cancer signatures. By flagging cases with elevated risk before overt tumor development, REDMOD raises the prospect of earlier intervention and potentially curative treatment, heralding a shift from reactive to predictive oncology.

Coupled with clinical expertise, the AI model is currently undergoing prospective clinical testing within the Artificial Intelligence for Pancreatic Cancer Early Detection (AI-PACED) study. This initiative evaluates how REDMOD-guided interpretation can be integrated into patient management, measuring outcomes such as early detection rates, false-positive frequency, and downstream clinical benefit. The prospective design and real-world setting serve to validate the utility of AI not just as a retrospective research tool but as a practical adjunct in routine clinical care.

This research is a flagship component of the Mayo Clinic’s Precure initiative, which aims to preempt disease through early biological detection and intervention, fundamentally reshaping the future of healthcare. It also aligns closely with Mayo Clinic’s Clinical Impact strategy focused on accelerating the translation of scientific discoveries into tangible patient benefits. Such institutional commitment underscores the transformative nature of this AI development and its consonance with the evolving landscape of precision medicine.

Importantly, the Mayo Clinic team has leveraged substantial funding and collaborative infrastructure, receiving support from the National Institutes of Health and distinguished philanthropic foundations. This robust backing facilitated the extensive multi-institutional datasets essential for rigorous training and validation of AI algorithms, preserving generalizability and minimizing overfitting. Such thorough experimental design is paramount to ensuring that AI tools meet the stringent standards required for clinical acceptance and regulatory approval.

The implications of REDMOD extend beyond pancreatic cancer. This approach exemplifies the power of integrating advanced imaging informatics and machine learning to reveal “invisible” disease phenotypes at presymptomatic stages. Other malignancies and chronic diseases with heterogeneous tissue alterations may similarly benefit from AI analytics applied to routine imaging, heralding a new era where artificial intelligence fundamentally enhances radiologic interpretation.

In conclusion, Mayo Clinic’s REDMOD represents a landmark leap toward transforming pancreatic cancer from a disease diagnosed too late for cure into one detectable in its silent inception. By extracting rich radiomic data imperceptible to humans, this AI model offers a new window into early disease states. As clinical testing progresses, the hope is that this technology will integrate seamlessly with standard care, enabling timely intervention and ultimately improving survival outcomes in what has historically been one of the most challenging cancers to diagnose and treat.

Subject of Research: Early detection of pancreatic cancer using AI-enhanced radiomics on routine CT scans

Article Title: Next-generation AI for visually occult pancreatic cancer detection in a low-prevalence setting with longitudinal stability and multi-institutional generalisability

News Publication Date: April 22, 2026

Web References:

• Mayo Clinic: https://www.mayoclinic.org/about-mayo-clinic
• Pancreatic cancer information: https://www.mayoclinic.org/diseases-conditions/pancreatic-cancer/symptoms-causes/syc-20355421
• Published study: https://gut.bmj.com/content/early/2026/04/22/gutjnl-2025-337266

References: Refer to the Gut journal publication: “Next-generation AI for visually occult pancreatic cancer detection in a low-prevalence setting with longitudinal stability and multi-institutional generalisability”

Image Credits: Not provided

Keywords: Pancreatic cancer, early detection, artificial intelligence, radiomics, CT imaging, AI-PACED study, Mayo Clinic, radiology, oncology, machine learning, precision medicine, diagnostic innovation

Pancreatic ductal adenocarcinoma remains one of the deadliest cancers, largely because it is customarily diagnosed at an advanced stage when therapeutic interventions offer minimal benefit. The aggressive nature of PDAC combined with its clinical silence contributes to dismal survival statistics, with many cases identified only after metastasis. Traditional computed tomography (CT) scans and clinical evaluations, despite their utility in many oncologic contexts, often fail to reveal the subtle microarchitectural tissue changes that herald the earliest cancerous transformations within the pancreas. This has fostered an urgent need for novel detection modalities capable of revealing these invisible early signs.

Addressing this critical gap, the REDMOD framework harnesses the power of radiomics—the extraction and analysis of complex quantitative features from medical images—combined with automated pancreas segmentation. This segmentation enables precise delineation of the pancreatic borders from surrounding tissues without manual oversight, mitigating risks associated with human error and variability. Such technical sophistication ensures that the AI examines consistent regions with high fidelity across diverse imaging datasets, a necessity for reliable early detection in a real-world clinical context.

To evaluate its clinical validity, the AI was retrospectively applied to abdominal CT scans from 219 patients initially deemed disease-free by radiologists but who were subsequently diagnosed with PDAC. These scans spanned time intervals extending up to three years prior to official diagnosis. Impressively, REDMOD identified pre-clinical malignant signatures an average of 475 days—approximately 15 months—before the clinical diagnosis was made. Notably, nearly two-thirds of these cancers were localized to the pancreatic head, an area notoriously challenging to assess. This discovery underscores a significant temporal window during which early intervention could substantially alter patient outcomes.

Comparison with a large control group comprising 1,243 age-, sex-, and scan-date matched individuals who remained PDAC-free for over three years highlighted REDMOD’s specificity. The model accurately recognized over 81% of cases as negative for cancer in an independent multicenter cohort and demonstrated 87.5% accuracy in a publicly available NIH dataset. Such high specificity is crucial in minimizing false positives that can lead to anxiety and unnecessary medical procedures. The consistency of REDMOD’s output was further bolstered by repeat scans from the same patients, which yielded 90 to 92% concordance in detecting early malignant signatures months apart, illustrating the AI’s longitudinal reliability.

Perhaps most compelling is REDMOD’s performance relative to highly experienced radiologists. The AI achieved a sensitivity of 73% in detecting early-stage PDAC changes—nearly double the 39% sensitivity attributed to human experts. This gap widened dramatically for cases detected over two years before clinical diagnosis, with REDMOD maintaining a 68% accuracy while radiologist detection fell to only 23%. These findings challenge the current clinical reliance on human interpretation alone and advocate for AI integration to capture otherwise invisible radiological cues.

Despite these promising results, the lead researchers cautiously note that further validation in prospective, high-risk patient groups remains imperative before widespread clinical adoption. Patients exhibiting symptoms such as unexpected weight loss or recent-onset diabetes—conditions often associated with increased PDAC risk—may benefit most from such AI surveillance. Moreover, while this study benefitted from multi-institutional data enhancing its generalizability, the participant demographics lacked ethnic diversity, signaling an area for expansion in future research.

At its core, the REDMOD framework represents a convergence of advanced computational imaging analysis and clinical oncology. The system’s fully automated design eliminates the bottleneck of manual image segmentation, accelerating processing and reducing inter-operator variability. Beyond pancreatic cancer, such methodologies herald a new era of radiological precision medicine, wherein hidden oncologic processes can be unmasked well before they manifest clinically or morphologically to the human eye.

The implications of this research extend into health economics and patient quality of life. Modeling indicates that increasing early-stage, localized PDAC detection from 10% to 50% could more than double survival rates, illustrating the profound influence of diagnostic timing. In a disease where late diagnosis is the norm and effective treatment options are limited, the ability to non-invasively detect cancer over a year in advance transforms the clinical landscape, offering hope where few options previously existed.

In summary, REDMOD embodies a significant leap toward proactive pancreatic cancer detection. By unveiling the “invisible” textures of early cellular malignant transformation within routine CT scans, it empowers clinicians with unprecedented foresight. Although prospective trials and validation in diverse populations are essential next steps, this research lays the foundation for AI-enhanced diagnostic pathways that could save countless lives and fundamentally change the prognosis of a devastating cancer.

Subject of Research: People

Article Title: Next-generation AI for visually occult pancreatic cancer detection in a low-prevalence setting with longitudinal stability and multi-institutional generalisability

News Publication Date: 28-Apr-2026

Web References: 10.1136/gutjnl-2025-337266

Keywords

Pancreatic cancer, Artificial intelligence, Imaging, Radiology, Early detection, Pancreatic ductal adenocarcinoma, Radiomics, Computed tomography, Automated segmentation, Medical imaging analysis

Pancreatic cancer has long been recognized as one of the deadliest malignancies, boasting a dismal five-year survival rate of only 13%. Its notorious reputation stems from its late diagnosis, often characterized as a “silent killer,” alongside its ability to swiftly adapt to therapeutic interventions. The current findings reveal that a substantial portion of the tumor cells leverage ecDNA to facilitate a rapid response to environmental pressures, essentially equipping them with a robust mechanism for survival and growth.

The researchers unearthed that many pancreatic cancer cells harbor multiple copies of vital oncogenes, such as MYC, situated on circular pieces of DNA that exist independently of chromosomes. This unique structure, ecDNA, which is capable of floating freely within the cell nucleus, enables these tumor cells to exhibit dynamic gene expression, thereby allowing them to modify their morphology and endure under unfavorable conditions. This discovery shows that pancreatic cancer cells employ ecDNA as a critical tool in their evolutionary toolkit, showcasing the complexity and adaptability inherent in these cancer cells.

Co-corresponding author Peter Bailey from the Botton-Champalimaud Centre elaborated on the findings, emphasizing the lethality of pancreatic cancer and its stealthy nature. The ability of these tumor cells to “shape-shift” under stress is underscored by the presence of ecDNA, marking a pivotal shift in understanding how some tumors can gain a survival advantage, especially in highly stressful environments. Notably, the researchers observed that this genomic feature was prevalent in pancreatic tumors, particularly for oncogenes like MYC, thus painting a broader picture of the tumor’s adaptable behavior.

The significant increase in variability of MYC copy number was noted particularly when MYC resided on ecDNA. The researchers found that certain cells possessed numerous additional copies of MYC, dramatically boosting their growth potential under specific conditions. This phenomenon illustrates a ‘bet-hedging’ strategy, where diverse populations within the tumor carry differing amounts of MYC. Some cells thrive under high MYC expression, while others with lower levels may fare better in alternative environments. This intricate balancing act among the populations encapsulates the profound intratumor heterogeneity that defines pancreatic cancer.

A pivotal methodology employed in this study involved the use of organoids—three-dimensional miniaturized models of pancreatic tumors derived directly from patients with early-stage pancreatic cancer. These organoids authentically replicate the genetic landscape of the original tumors, providing an invaluable platform for cancer research. Unlike traditional approaches that manipulate ecDNA artificially, studying these organoids enables researchers to observe genuine ecDNA variations and their effects within the tumor environment.

In exploring how ecDNA contributes to cancer cell adaptation, the researchers subjected the patient-derived organoids to a controlled environment where they removed critical growth signals, such as WNT factors. This experimental design illuminated the ability of organoids carrying MYC on ecDNA to become more autonomous, reducing their reliance on external growth signals for survival. The findings were particularly significant, as they reveal the adaptability of cells that harbor ecDNA—bolstering their self-sufficiency amidst hostile conditions.

As the study progressed, it became evident that high MYC levels correlated with substantial changes in tumor cell morphology and behavior. Increased ecDNA levels caused the cells to adopt more aggressive, solid structures, often at the expense of their gland-like organization. These observations led researchers to suggest that the emergence of ecDNA allows for rapid genomic adaptations, prompting cells to respond effectively to fluctuating environmental pressures by adjusting their morphologies and functional dynamics.

In a remarkable finding, the researchers pointed out that ecDNA-endowed copies of MYC can appear and disappear with astonishing rapidity depending on external stimuli. For instance, in circumstances of acute growth factor deprivation, those cancer cells could significantly ramp up MYC expression, securing their survival advantage. Conversely, under less stressful conditions, these cells might selectively shed some of the excess ecDNA to minimize the potential risks associated with high MYC levels, which can lead to DNA damage.

The implications of these findings on the therapeutic landscape for pancreatic cancer could be vast. Although ecDNA appeared in approximately 15% of the samples analyzed in this study, this subset may represent a particularly aggressive faction of tumors that possess heightened resistance to established therapies. Consequently, identifying or targeting ecDNA could herald new opportunities for treatment strategies aimed at enhancing patient outcomes.

A therapeutic approach could involve pushing these cancer cells to overload on MYC expression, thereby inducing an unmanageable level of DNA damage. Alternatively, blocking the molecular pathways that sustain these ecDNA structures might lead to their degradation, forcing tumor cells to lose their genetic advantages. However, researchers caution that these strategies remain preliminary at this stage, as the dual nature of ecDNA poses significant challenges; while it facilitates rapid adaptation, it comes with the metabolic cost of maintaining extra genetic material.

The research broadens our comprehension of genomic plasticity, dispelling the notion that our genomes are static and immutable. The discovery that WNT signaling can directly influence DNA architecture was unexpected and highlights the dynamism with which tumors can evolve and adapt to their microenvironments. The implications are particularly crucial given the projected increase in pancreatic cancer incidence in upcoming years, signaling an urgent need for innovative interventions that target these genetic features to enhance treatment susceptibility.

As we delve deeper into the nature of pancreatic cancer, the multifaceted role of ecDNA emerges as a formidable topic that merits further exploration. The potential for targeted therapies that exploit the vulnerabilities linked to ecDNA could pave the way for advancements in treatment protocols. Ultimately, continued research in this area will be instrumental in uncovering new strategies to combat a disease that has historically posed significant therapeutic challenges.

This research underscores the importance of continuing to investigate complex genetic features that influence cancer behavior and treatment outcomes. By understanding the nuanced dynamics of ecDNA and its implications in intratumor heterogeneity, scientists may unlock new pathways for intervention, ultimately improving the prognosis for patients with one of the most formidable adversaries in oncology.

Subject of Research: Pancreatic cancer and its relationship with extrachromosomal DNA.
Article Title: MYC ecDNA promotes intratumor heterogeneity and plasticity in PDAC.
News Publication Date: 12-Mar-2025.
Web References: Link to Journal
References: Nature.
Image Credits: Vinzenzo Corbo Lab.
Keywords: pancreatic cancer, extrachromosomal DNA, MYC, tumor heterogeneity, therapy resistance, genomic plasticity.

Current screening methods for pancreatic cancer, while informative, often fall short when it comes to early detection. Traditional biomarkers, such as the carbohydrate antigen 19-9 (CA 19-9), are valuable for prognosis but lack the sensitivity required for identifying pancreatic cancer in its incipient stages. PAC-MANN fills this crucial gap by providing a tool that can pinpoint cancer-related activity in the bloodstream, thereby facilitating earlier diagnosis and optimizing treatment outcomes. The significance of this breakthrough is amplified by the fact that early detection is typically correlated with a more favorable prognosis, making it imperative for the medical field to adopt more sensitive screening methodologies.

Dr. Jared Fischer, a key scientist at the OHSU Knight Cancer Institute’s Cancer Early Detection Advanced Research Center (CEDAR), emphasizes the critical need for such a test. He states, “The problem with pancreatic cancer is that we often catch it too late.” Fischer and his colleagues have dedicated their research efforts to creating a more robust and user-friendly method for detecting pancreatic cancer, focusing on the practicality and accessibility of the PAC-MANN test. This endeavor aims to empower clinicians with tools that offer timely insights into patient health, particularly for those at an elevated risk for this devastating disease.

The development process for PAC-MANN involved analyzing blood samples from a diverse cohort of 350 patients, who approximately ranged from those diagnosed with pancreatic cancer, to those identified as high-risk candidates, as well as a control group. The researchers meticulously targeted specific proteins, particularly proteases, which exhibit increased activity in the presence of PDAC. This understanding enabled them to refine the test to recognize the protein markers associated with pancreatic cancer, thereby paving the way for a reliable and efficient diagnostic tool that works outside the confines of specialized laboratories.

In clinical evaluations, the PAC-MANN test demonstrated an impressive ability to distinguish between patients with pancreatic cancer and those without—achieving accuracy rates of 98%. This capability signifies not only a potential paradigm shift in diagnostic procedures but also a means to enhance patient monitoring throughout their treatment regimens, offering real-time adjustments based on protease activity levels. Such responsiveness allows clinicians to tailor therapeutic strategies that align with an individual patient’s specific needs, fundamentally transforming personalized medicine in oncology.

Research findings indicate that the PAC-MANN test doesn’t merely excel at initial detection; it also holds promise in tracking treatment efficacy over the course of a patient’s therapy. Following surgical interventions, a decrease in protease activity was observed, suggesting that this testing method could render critical insights into how well a patient’s treatment is progressing. As Dr. Fischer highlights, this capability to assess response in real time could contribute significantly to improving clinical decision-making and optimizing patient outcomes.

The accessibility of the PAC-MANN test is a notable aspect of its design. Unlike traditional methods, which often necessitate larger volumes of blood and complex procedures, PAC-MANN requires only 8 microliters of blood, rendering it a less invasive approach for screening potential pancreatic cancer cases. The promise of rapid results—within approximately 45 minutes—and the astonishingly low cost of less than a penny per sample enhances its appeal for widespread clinical adoption, particularly in underserved communities that often lack access to traditional testing modalities.

Looking ahead, there are plans for additional clinical trials, specifically tailored towards patients at high risk for developing pancreatic cancer. These investigator-initiated trials are grounded in a collaborative effort involving the OHSU Knight Cancer Institute and CEDAR, reflecting a commitment to validating the efficacy of the PAC-MANN test in real-world settings. If successful, these trials will further establish the test’s utility and effectiveness, potentially integrating PAC-MANN into routine clinical practice for pancreatic cancer screening and management.

Dr. Montoya Mira, a research engineer and co-author of the study, notes the aim of developing an economically viable and efficient testing method that transcends the limitations of existing practices. He envisions a future where PAC-MANN can be utilized not only in urban hospitals but also in rural settings where logistical constraints could jeopardize early cancer detection. This vision encapsulates the core mission of the researchers—to democratize access to essential healthcare tools that can save lives and improve outcomes for patients facing pancreatic cancer.

In summary, the PAC-MANN test represents a landmark achievement in the ongoing battle against pancreatic cancer. By enabling earlier detection and facilitating the monitoring of treatment efficacy, this innovative blood test could fundamentally change the landscape of care for patients suffering from one of the most lethal cancers. The potential impact of such a tool echoes through the echoes of the oncology community, providing a glimmer of hope in an often grim scenario characterized by late-stage diagnoses and limited therapeutic options. As the research moves forward, the excitement builds around the possibility of a future where pancreatic cancer is caught in its earliest stages, significantly enhancing survival chances for those affected.

Subject of Research: Human tissue samples
Article Title: PAC-MANN in the pancreas: Blood test for early pancreatic cancer shows promise in trial
News Publication Date: 12-Feb-2025
Web References: http://dx.doi.org/10.1126/scitranslmed.adq3110
References:
Image Credits: OHSU/Christine Torres Hicks

Keywords: Pancreatic cancer, blood test, protease activity, early detection, oncology, PAC-MANN, OHSU, treatment monitoring