CUP is not a single disease but a clinical condition in which cancer has spread while its original site remains hidden. In many patients, doctors use imaging, pathology, immunohistochemistry and molecular tests to search for the primary tumour. Even after extensive investigation, however, the source may remain unknown. This creates a major treatment challenge because cancer therapies are often selected according to the tissue in which a tumour originated. A cancer that began in the lung, breast, bowel or pancreas may respond to very different drugs, yet CUP can prevent clinicians from making that distinction with confidence.

The CUP-COMP trial is focused on a technology known as liquid biopsy. Instead of requiring a tumour sample obtained through surgery or an invasive biopsy, liquid biopsy analyses biological material circulating in the blood. Tumour cells can release fragments of DNA into the bloodstream, including circulating tumour DNA, or ctDNA. These fragments may contain mutations, copy-number changes and other molecular abnormalities that reflect the biology of the cancer. By sequencing this material, researchers can search for patterns that may help classify a tumour, reveal potentially targetable alterations or indicate how the disease is changing over time.

The central question is not simply whether such a test can produce a technically impressive molecular profile. The trial is designed to investigate whether blood-based profiling can be delivered reliably and usefully in routine clinical pathways for patients with CUP. That distinction is crucial. A test may work in a laboratory but prove difficult to implement at a national scale if blood samples arrive too late, contain too little tumour-derived DNA, fail quality-control checks or produce results that clinicians cannot interpret within the time available for treatment decisions.

A prospective design allows the researchers to evaluate these issues as they occur, rather than relying only on stored samples or retrospective records. Patients can be followed through the process of consent, blood collection, sample transport, laboratory analysis and clinical reporting. This approach can reveal where delays and failures arise and whether the information generated is available at a moment when it might influence patient care. It also provides a framework for measuring how often blood samples yield an interpretable molecular profile and how consistently testing can be integrated across participating centres.

Technically, the approach may combine several layers of genomic information. DNA sequencing can identify mutations in genes that drive tumour growth or create vulnerabilities to targeted drugs. Copy-number analysis can detect gains and losses of DNA segments, while broader molecular signatures may offer clues about tumour lineage. Some platforms can also estimate the fraction of DNA in a blood sample that originates from the tumour, a measurement known as tumour fraction. When this fraction is low, a negative result may not mean that a mutation is absent; it may simply indicate that the test did not receive enough tumour-derived material to detect it.

That limitation is particularly important in CUP, where disease biology can vary widely and metastatic deposits may release unequal amounts of DNA into the circulation. Tumour burden, the location of metastases, treatment exposure and the biology of individual cancers can all affect ctDNA levels. Blood-based profiling therefore does not eliminate the need for clinical assessment, imaging or tissue pathology. Instead, it is being evaluated as an additional source of evidence that could complement established diagnostic methods and potentially reduce the time required to obtain molecular information.

The precision-medicine element of CUP-COMP reflects a broader shift in cancer treatment. Rather than assigning therapy solely according to the organ where a tumour started, oncologists increasingly seek molecular features that can be targeted directly. Alterations in genes involved in DNA repair, cell signalling or immune regulation may appear across cancers from different organs. If these abnormalities are detected in a patient with CUP, they could provide a rationale for considering a targeted therapy or an immunotherapy, although the clinical value of any proposed treatment must still be assessed through evidence, eligibility criteria and multidisciplinary review.

The study also addresses an important question of equity and scalability. Advanced molecular testing is not useful if it is available only at specialist institutions or to patients who can access highly centralised services. A national trial can test whether samples collected in different hospitals can be processed through a coordinated system and whether results can be returned in a consistent format. The findings may help establish the logistical requirements for wider adoption, including laboratory capacity, data interpretation, reporting standards and communication between molecular scientists and treating teams.

For patients facing a diagnosis in which the primary tumour cannot be found, the promise of a blood-based test is therefore measured not only in scientific novelty but in speed, accessibility and clinical clarity. CUP-COMP is evaluating whether molecular information can be obtained from a relatively simple blood draw and incorporated into real-world decision-making. Its significance will ultimately depend on whether the approach produces dependable results, identifies actionable biology and fits the demanding timelines of cancer care. By testing those questions prospectively, the study may help determine whether liquid biopsy can become a practical component of precision medicine for one of oncology’s most uncertain and difficult diagnoses.

Subject of Research: Blood-based molecular profiling and precision medicine for patients with Cancer of Unknown Primary (CUP).

Article Title: A prospective national precision medicine trial evaluating the feasibility of blood-based molecular profiling in patients with Cancer of Unknown Primary (CUP-COMP).

Article References: Conway, AM., Robinson, M., Concannon, M. et al. A prospective national precision medicine trial evaluating the feasibility of blood-based molecular profiling in patients with Cancer of Unknown Primary (CUP-COMP). Br J Cancer (2026). https://doi.org/10.1038/s41416-026-03519-6

Image Credits: AI Generated

DOI: 10.1038/s41416-026-03519-6

Keywords: Cancer of Unknown Primary, CUP, liquid biopsy, circulating tumour DNA, molecular profiling, precision medicine, cancer genomics, oncology, blood-based testing.

The novel device developed by Professor Chi Ming Che, Zhou Guangzhao Professor in Natural Sciences and Chair Professor of Chemistry at HKU, in collaboration with Dr. Wei Liu, represents a paradigm shift in the approach to cancer diagnostics. Bridging synthetic chemistry with cutting-edge artificial intelligence, this portable instrument offers a rapid, straightforward, and user-friendly cancer risk assessment that eschews the need for tissue biopsies or complex laboratory infrastructure. This innovation was recently lauded with the prestigious Gold Medal and Congratulations of the Jury at the 51st International Exhibition of Inventions of Geneva (2026), underscoring its scientific significance and potential to transform public health monitoring on a global scale.

At the heart of this technological marvel lies a unique class of luminescent metal complexes synthesized under Professor Che’s guidance. These metal complexes possess an extraordinary affinity for damaged DNA sites — particularly mismatches — which often serve as molecular hallmarks of oncogenic processes. Unlike conventional dyes or probes, these complexes undergo pronounced changes in their photoluminescent properties upon binding to compromised DNA strands, generating an optical signal of remarkable sensitivity and specificity. This luminescence phenomenon is directly correlated with the extent of DNA damage, allowing for quantitative assessment of cancer-related molecular aberrations without cumbersome sample preparation or specialized labeling.

To capture and interpret these delicate optical signals, the research team developed a miniaturized, high-precision spectrometer engineered by Dr. Wei Liu. This spectrometer operates seamlessly within the handheld device, detecting fluctuations in emission spectra triggered by the DNA-bound luminescent probes. Crucially, the raw spectroscopic data is fed into an advanced artificial intelligence engine that executes sophisticated pattern recognition and machine learning algorithms. This AI component distills complex optical signatures into clinically actionable insights, enhancing both the accuracy and speed of cancer risk prediction. The marriage of molecular sensing with AI-powered analytics heralds a new era where diagnostic precision meets digital efficiency.

Designed with portability and accessibility in mind, the device empowers individuals to conduct self-administered cancer risk screenings using merely a saliva sample, circumventing the discomfort and risks associated with invasive tissue biopsies. The entire detection process unfolds within ten minutes, facilitated via an intuitive mobile application interface that guides users through sample collection, analysis, and interpretation of results. This democratization of cancer screening holds immense promise, particularly for high-risk populations such as individuals with familial cancer histories or patients under continuous post-treatment surveillance, who require frequent and hassle-free monitoring.

Professor Che emphasizes that while this groundbreaking tool is not intended to supplant established clinical diagnostic procedures, it serves as a potent auxiliary platform for rapid detection and longitudinal tracking. Preliminary clinical investigations involving patients diagnosed with breast cancer and nasopharyngeal carcinoma have yielded compelling evidence of the device’s capability to discriminate effectively between patients afflicted by malignancy and healthy individuals. These encouraging findings lay the groundwork for expansive validation efforts, as the HKU research team presently collaborates closely with oncologists from multiple hospitals to assess the technology’s efficacy across a diverse array of cancer types and patient cohorts.

Beyond its clinical applications, the technology exemplifies the power of interdisciplinary innovation — uniting the realms of synthetic chemistry, optical physics, and artificial intelligence into a harmonious diagnostic ecosystem. The luminescent metal complexes, a novel chemical entity crafted through meticulous molecular design, underscore the potential of chemical biology to yield tools that decipher complex biological phenomena at a molecular level. Meanwhile, AI’s capacity to parse multifaceted data patterns in real-time offers unprecedented advantages in translating these molecular events into reliable health indicators.

The societal implications of this development are profound. Cancer imposes staggering costs not only in lives lost but also in economic and social hardships. Early detection and continuous monitoring reduce these burdens by enabling timely interventions that improve prognoses and conserve healthcare resources. By delivering an easily deployable, low-cost, and scalable technology, this device could markedly enhance screening coverage, especially in underserved or resource-limited regions where traditional diagnostic infrastructure is scarce.

Moreover, the technology aligns with broader trends in personalized and precision medicine, where diagnostic tools tailor healthcare responses to individual molecular profiles. Its ability to detect subtle DNA damage signatures non-invasively dovetails with efforts to shift cancer care upstream — focusing on prevention, early interception, and personalized risk stratification. As the device integrates seamlessly with digital health platforms, it can potentially interface with telemedicine services, further extending its reach and impact.

In essence, this AI-integrated optical sensing device not only embodies a leap forward in cancer diagnostics but also illustrates a compelling blueprint for the next generation of biomedical innovations: compact, intelligent, and patient-centric technologies designed to empower individuals and enhance public health outcomes. The convergence of chemical ingenuity and artificial intelligence opens new vistas for detecting and understanding disease processes in ways previously unattainable, bringing us closer to a future where cancer detection is swift, safe, and universally accessible.

The University of Hong Kong and the Laboratory for Synthetic Chemistry and Chemical Biology Limited (LSCCB) continue to spearhead this ambitious initiative, striving to translate laboratory breakthroughs into tangible clinical benefits. Their ongoing collaborations with medical practitioners and commitment to rigorous validation promise to refine and optimize this technology for broader clinical deployment. With further development and integration, this innovative device could become an indispensable tool in the global fight against cancer, exemplifying how scientific excellence can be harnessed to achieve meaningful societal impact.

For inquiries related to this pioneering research, contact the Office of Vice-President and Pro-Vice-Chancellor (Research) at The University of Hong Kong, or Ms. Esther YIU via telephone or email.

Subject of Research: Development of a portable AI-enabled optical sensing device for rapid, non-invasive cancer risk detection using saliva samples.

Article Title: AI-Powered Optical Device Enables Rapid, Non-Invasive Cancer Risk Screening via Saliva Analysis

News Publication Date: Not specified

Web References: Not specified

References: Not specified

Image Credits: The University of Hong Kong

Keywords: Cancer detection, non-invasive diagnostics, optical sensing, luminescent metal complexes, artificial intelligence, saliva-based screening, biosensors, molecular diagnostics, digital health, early cancer screening

Liver cancer, particularly HCC, often emerges against a backdrop of chronic HBV infection and subsequent liver damage, including cirrhosis. Despite advances in medical imaging and serum biomarkers, catching HCC at an early, treatable stage has proved elusive. The promise of extracellular vesicles as carriers of disease-specific molecular signatures opens new frontiers. These nanometer-sized vesicles, secreted by cells into bodily fluids, encapsulate a rich cargo of RNAs, proteins, and lipids reflective of their cellular origin, thus serving as a “liquid biopsy” without the invasiveness of traditional tissue sampling.

In this comprehensive study, serum EVs were isolated from a cohort consisting of healthy controls, chronic hepatitis B (CHB) patients, liver cirrhosis patients, hepatocellular adenoma patients, and those diagnosed with HCC. The use of ultracentrifugation ensured high-purity vesicle isolation, while transmission electron microscopy, nanoparticle tracking analysis, and Western blotting confirmed the isolated EVs’ identity and purity. This rigorous validation underpins the credibility of subsequent molecular analyses.

High-throughput transcriptome sequencing was employed to profile RNA content within EVs from each clinical group, enabling systematic comparisons of lncRNA expression associated with disease progression. The study identified an array of 133 lncRNAs demonstrating significant differential expression specifically in the HCC group, underscoring their potential as biomarkers uniquely linked to malignant transformation in HBV-related liver disease.

The analytical framework extended beyond mere identification. Through multi-step screening and time-series analysis, the researchers pinpointed 10 core lncRNAs closely correlated with HCC progression. These lncRNAs exhibit dynamic expression changes aligning with clinical stages, suggesting their active involvement in the tumorigenic process rather than passive association. Such specificity is key to their potential deployment in diagnostic applications.

Diving deeper into molecular mechanisms, the authors constructed a complex lncRNA-miRNA-mRNA regulatory network encompassing 62 nodes and 68 interactions. This network sheds light on the layered post-transcriptional regulation and cross-talk among diverse RNA species. It highlights how lncRNAs may act as competing endogenous RNAs (ceRNAs), modulating miRNA availability and downstream mRNA expression, thereby influencing cellular pathways relevant to tumor growth and survival.

Functional enrichment analyses provided compelling hints about the biological processes modulated by these lncRNAs. The implicated pathways include critical aspects of cell proliferation regulation, transmembrane ion transport, cytosolic and plasma membrane localization, protein binding interactions, and vital signaling cascades such as autophagy and the mitogen-activated protein kinase (MAPK) pathway. These findings reveal the multifaceted impact of EV-derived lncRNAs on cellular homeostasis and oncogenic signaling networks.

Protein-protein interaction (PPI) network analysis further distilled the hub genes within this regulatory landscape, identifying 10 key genes including NTRK2 and KCNJ10. These hub genes likely serve as pivotal nodes mediating cross-talk within the signaling circuitry, rendering them potential targets for therapeutic intervention or biomarker validation.

To ensure robustness, the study validated the expression patterns of core lncRNAs and their downstream genes using an independent plasma cohort. The consistency observed across distinct patient populations strengthens the case for these molecules as reproducible biomarkers with clinical diagnostic value, potentially enabling real-time monitoring of disease progression via minimally invasive blood tests.

The implications of these findings are profound. By elucidating a set of HCC-specific lncRNA biomarkers packaged within extracellular vesicles, the study pioneers a paradigm enabling clinicians to leverage liquid biopsy techniques for early detection of liver cancer in high-risk HBV-infected individuals. Such breakthroughs promise to enhance prognosis by facilitating timely therapeutic interventions and personalized treatment strategies.

Moreover, the mechanistic insights into EV lncRNA-mediated regulatory networks enhance our understanding of tumor biology, possibly unveiling novel therapeutic avenues aimed at disrupting pathological signaling cascades in HCC. Targeting these EV-associated lncRNAs or their interacting partners could augment current treatment modalities and improve patient outcomes.

This research underscores the formidable potential of integrating advanced molecular profiling with cutting-edge bioinformatic analyses to decode the complexities of cancer progression. The marriage of transcriptomics, network biology, and clinical validation exemplifies a holistic approach that could be adapted to other malignancies where EV-derived molecules serve as biomarkers and mediators.

As the scientific community continues to grapple with liver cancer’s global toll, discoveries like these mark a critical stepping stone towards mitigating disease burden through early, precise, and non-invasive diagnosis. The promise of EV-derived lncRNAs heralds a new era where liquid biopsies transcend experimental status to become standard clinical tools.

Future research will likely explore how these EV-lncRNA signatures interact with the immune microenvironment, influence metastatic potential, and respond to therapeutic pressures. Longitudinal studies across larger cohorts will also be essential to verify clinical utility and refine biomarker panels for widespread screening initiatives.

In conclusion, this pioneering investigation charts a sophisticated molecular atlas of EV-derived lncRNAs linked to HBV-related HCC progression. It not only illuminates key biological pathways modulated during hepatocarcinogenesis but also lays the groundwork for transformative liquid biopsy-based diagnostic platforms. As such, it offers renewed hope for millions threatened by liver cancer worldwide.

Subject of Research: Characteristics and mechanistic roles of extracellular vesicle-derived long non-coding RNAs during HBV-related hepatocellular carcinoma progression.

Article Title: Characteristics of extracellular vesicle-derived lncRNAs during the progression of HBV-related hepatocellular carcinoma

Article References:
Ma, Y., Lou, C., liang, J. et al. Characteristics of extracellular vesicle-derived lncRNAs during the progression of HBV-related hepatocellular carcinoma. BMC Cancer 25, 1768 (2025). https://doi.org/10.1186/s12885-025-15237-y

Image Credits: Scienmag.com

DOI: 10.1186/s12885-025-15237-y (Published 14 November 2025)

This review, authored by Kelly, Samarska, Ramaekers, and colleagues, offers an unprecedented consolidation of data from 46 articles that collectively evaluated 105 unique urinary biomarkers. Spanning metabolites, proteins, microRNAs (miRNAs), DNA methylation markers, and other molecular entities, the study scrutinizes their diagnostic capabilities against control groups, emphasizing rigorous criteria such as area under the curve (AUC) values to gauge clinical utility. Notably, the systematic approach leveraged databases including PubMed, Scopus, and Web of Science, supplementing with cross-referenced literature to compile a robust and current repository of evidence.

At the heart of this inquiry lies the tension between the abundant preliminary findings and the stark reality of clinical translation failure. Biomarkers such as urinary proteins AQP1 (aquaporin-1) and PLIN2 (perilipin 2) emerged as particularly promising. These proteins are implicated in tumor-driven metabolic rerouting, offering a window into the dysregulated biochemical landscapes distinctive of RCC. Their diagnostic power, quantified by AUC scores exceeding 0.80 in several studies, underscores their potential to revolutionize non-invasive RCC detection protocols.

Equally compelling are dysregulated energy metabolism markers identified among urinary metabolites. RCC tumorigenesis involves metabolic reprogramming, redirecting normal renal cellular processes toward enhanced glycolysis and altered lipid metabolism. These biochemical signatures, detectable in urine, provide a functional snapshot that complements anatomical imaging. Metabolomic profiling, therefore, may not only differentiate malignant from benign lesions but also illuminate tumor aggressiveness and subtype-specific characteristics.

MicroRNAs—small, non-coding RNAs that regulate gene expression—further enrich the diagnostic tapestry. Among these, miR-122-5p, miR-15a, and miR-30c showed notable differential expression patterns consistently across patient cohorts. These miRNAs modulate pathways integral to tumor proliferation and apoptosis, rendering them attractive biomarker candidates. Still, the evidence compels cautious optimism, as validation studies are sparse and often limited by sample size and heterogeneity.

Intriguingly, the study illuminates that composite multi-biomarker panels frequently outshine individual markers. This synergy principle harnesses the multifactorial nature of RCC pathophysiology, where a constellation of altered molecular signals can amplify diagnostic precision. Some panels integrated metabolites with proteins or miRNAs, reflecting a layered approach to capture the complexity of tumor biology through urinary assays. This strategy might bridge sensitivity and specificity gaps encountered when relying on single biomarkers alone.

Yet, a conspicuous gap persists in external validation—an indispensable step before clinical deployment. Many promising markers and panels were identified in isolated, small-scale cohorts, frequently lacking multi-center replication or independent review. The authors particularly highlight the necessity for standardized protocols and large prospective trials to authenticate the biomarkers’ diagnostic robustness and reproducibility across diverse populations.

The systematic review does not shy away from confronting methodological hurdles inherent in biomarker research. The included studies exhibited variability in sample collection, analytical techniques, and statistical handling, factors that complicate direct comparisons and meta-analyses. For instance, urine sample handling and storage conditions critically influence biomarker stability, and disparate assays range from mass spectrometry to PCR-based methods, each introducing unique biases and limitations.

A further challenge lies in the biological heterogeneity of RCC itself, which encompasses several histological subtypes with distinct molecular underpinnings. An ideal urinary biomarker or panel would therefore need to transcend this heterogeneity or incorporate subtype-specific markers to maximize clinical relevance. The current review points toward this necessity, encouraging future research to adopt nuanced stratification frameworks in study designs.

From a translational perspective, urinary biomarkers hold immense appeal due to their non-invasive nature, ease of repeated sampling, and potential for integration into routine clinical workflows. Unlike invasive biopsies or costly imaging techniques, urinary assays could serve as frontline screening tools or adjuncts for early detection, monitoring recurrence, and guiding therapeutic decisions. This potential mandates intensified efforts to surmount existing validation and standardization barriers.

Notably, the review adheres strictly to established reporting standards, utilizing the PRISMA guidelines for systematic reviews and the modified STROBE checklist to assess study bias. Such methodological rigor lends credibility to its conclusions and provides a transparent evaluation of the evidence quality, enabling clinicians and researchers to interpret findings within a robust framework.

The implications of this comprehensive review reach beyond RCC alone. They exemplify the broader biomarker development paradigm in oncology, highlighting the intricate balance between promising molecular discoveries and the pragmatic pathways toward clinical impact. The blueprint proposed underscores the indispensability of collaborative, multidisciplinary research efforts encompassing molecular biology, clinical oncology, bioinformatics, and biostatistics.

In summary, the 2025 systematic review in BMC Cancer heralds a pivotal moment for RCC diagnostics, synthesizing diverse molecular insights into a coherent narrative of potential and pitfalls. While urinary biomarkers such as AQP1, PLIN2, key metabolites, and select miRNAs exhibit commendable diagnostic promise, their clinical utility hinges on rigorous, large-scale validation and multi-marker panel refinement. This evolving field eagerly awaits future studies to cement these biomarkers’ roles, signaling a transformative advance toward non-invasive, precise RCC diagnosis.

Subject of Research: Diagnostic urinary biomarkers for renal cell carcinoma (RCC)

Article Title: Renal cell carcinoma detection: a systematic review in diagnostic urinary biomarkers

Article References: Kelly, J.F., Samarska, I.V., Ramaekers, B. et al. Renal cell carcinoma detection: a systematic review in diagnostic urinary biomarkers. BMC Cancer 25, 1672 (2025). https://doi.org/10.1186/s12885-025-14900-8

Image Credits: Scienmag.com

DOI: https://doi.org/10.1186/s12885-025-14900-8

Keywords: Renal cell carcinoma, RCC, urinary biomarkers, diagnostic biomarkers, metabolites, proteins, microRNAs, DNA methylation, multi-biomarker panels, non-invasive cancer detection

Perineural invasion, the process by which cancer cells infiltrate the nerves surrounding a tumor, is a critical biomarker linked to aggressive disease progression and poor outcomes in prostate cancer patients. Traditionally, detecting PNI has relied heavily on invasive biopsy procedures and pathological examination, which come with limitations in sensitivity and spatial accuracy. Addressing these challenges, the study pivots toward a non-invasive imaging strategy, leveraging mpMRI to capture intricate tumor heterogeneity and generate quantifiable biomarkers predictive of PNI.

The research incorporated a substantial retrospective cohort of 397 prostate cancer patients recruited from two distinct medical centers, enabling a robust evaluation across diverse clinical settings. These patients were segmented into three distinct groups: a training cohort of 173 individuals, an internal validation (in-vad) group of 74, and an external validation (ex-vad) cohort consisting of 150 patients. This structured division ensured rigorous model training and unbiased assessment of predictive capability.

At the core of this study lies the concept of habitat analysis, a technique devised to dissect the tumor microenvironment into spatially distinct “habitats” by integrating key mpMRI sequences — specifically, T2-weighted imaging (T2WI), diffusion-weighted imaging (DWI), and apparent diffusion coefficient (ADC) maps. This multiparametric fusion elucidates differing tissue characteristics within the tumor mass, such as variations in cellularity and extracellular matrix composition, that are otherwise imperceptible through conventional imaging alone.

Following habitat segmentation, the study applied a tailored deep learning framework to extract complex features from these subregions. Through a meticulous feature selection and filtration process, the researchers derived a composite score termed “radscore.” This radscore effectively encapsulates the heterogeneity-driven imaging biomarkers that correlate with the presence or absence of perineural invasion.

The investigative team constructed six predictive models to compare and optimize PNI detection. These included a purely clinical model based on conventional patient data, four habitat-specific models addressing individual tumor subregions, and a combined model merging clinical parameters with mpMRI-derived radiomics. The overarching goal was to ascertain which approach delivered the highest discriminative power.

Results from receiver operating characteristic (ROC) curve analysis were remarkable. The four habitat models exhibited formidable performance across all cohorts, with area under the curve (AUC) values ranging between 0.802 and 0.957. This high degree of accuracy underscores the utility of habitat-specific imaging markers in capturing the nuanced biology of perineural invasion.

The standalone clinical model, while informative, demonstrated relatively modest performance with AUCs of 0.832, 0.818, and 0.789 in the training, internal validation, and external validation sets, respectively. This gap highlighted the necessity of integrating imaging biomarkers with classic clinical data to achieve superior predictive fidelity.

Most notably, the combined model, which synthesized clinical data and habitat-based radiomic features, substantially outperformed all other models. In the training cohort, this integrated approach attained an exceptional AUC of 0.999, alongside near-perfect sensitivity and specificity of 1 and 0.955, respectively. Such precision indicates that the combined model could virtually eliminate false negatives and false positives, addressing a critical unmet need in prostate oncology diagnostics.

Further substantiating the clinical relevance, decision curve analysis (DCA) and clinical impact curve analysis demonstrated that the combined model offers tangible benefits in patient management decisions. This implies that incorporating this predictive tool in routine workflow could guide more personalized treatment planning, reduce unnecessary interventions, and potentially improve patient outcomes.

The significance of these findings is multi-dimensional. Firstly, this study exemplifies how quantitative imaging biomarkers, when paired with cutting-edge artificial intelligence, can transform subjective radiological evaluation into objective and reproducible diagnostics. The deployment of mpMRI-based habitat analysis offers a window into tumor microenvironment traits that are pivotal for understanding cancer aggressiveness.

Secondly, the use of deep learning pipelines enables the extraction of high-dimensional, non-linear features from imaging data that elude traditional radiomics and human interpretation. The radscore concept epitomizes this integration, proving that sophisticated computational methods can condense complex imaging phenotypes into actionable clinical predictors.

Moreover, this research sets a precedent for multi-institutional collaboration, validating the generalizability of imaging-based predictive models across heterogeneous patient populations and clinical settings. The use of an external validation cohort fortifies confidence that these findings are not confined to a single center’s imaging protocols or patient demographics.

Despite the triumphs, the investigators acknowledge that further prospective studies are warranted to evaluate the model’s performance in real-time clinical scenarios and to integrate it with emerging biomarkers such as genomic or proteomic data. Additionally, prospective trials could assess the impact of this predictive approach on therapeutic decision-making and long-term patient survival.

The promise of DL and habitat analysis also extends beyond prostate cancer, potentially catalyzing analogous advances in other solid tumors where perineural invasion and tumor heterogeneity profoundly influence prognosis. As imaging technology and computational models continue to evolve, such integrated tools will become indispensable in precision oncology.

In essence, this pioneering study illuminates a path toward non-invasive, accurate, and clinically actionable prediction of perineural invasion in prostate cancer. The alignment of multiparametric MRI, habitat analysis, and deep learning heralds a new era of imaging biomarker discovery, promising to enhance diagnostic confidence and ultimately reshape patient care paradigms in urologic oncology.

Subject of Research: Prediction of perineural invasion in prostate cancer using multiparametric MRI-based habitat analysis and deep learning.

Article Title: A novel MRI-based habitat analysis and deep learning for predicting perineural invasion in prostate cancer: a two-center study

Article References:
Deng, S., Huang, D., Han, X. et al. A novel MRI-based habitat analysis and deep learning for predicting perineural invasion in prostate cancer: a two-center study. BMC Cancer 25, 1367 (2025). https://doi.org/10.1186/s12885-025-14759-9

Image Credits: Scienmag.com

DOI: https://doi.org/10.1186/s12885-025-14759-9

The traditional approach to tumor diagnosis largely depends on tissue biopsies and histological examination—methods that can be invasive, risky, and sometimes inconclusive. This is particularly true for brain tumors, where surgical sampling can carry significant risks. The new crossNN model bypasses these challenges by focusing on the tumor’s epigenome—the collection of chemical modifications that regulate gene expression without altering the underlying DNA sequence. These epigenetic modifications act as molecular fingerprints unique to each tumor type, enabling precise identification and classification.

Epigenetic landscapes contain hundreds of thousands of modifications that switch genes on or off, creating patterns that are as unique to tumors as fingerprints are to individuals. By harnessing these complex patterns, the AI model can accurately differentiate between more than 170 types of tumors originating from various organs. Remarkably, the model achieves 99.1 percent accuracy in brain tumor classification and 97.8 percent across all tumor types, outperforming previous AI approaches in oncology diagnostics.

What sets the crossNN model apart is its foundation on a relatively simple neural network architecture, making the AI both highly explainable and traceable—a significant improvement over many “black-box” AI systems. This means clinicians and researchers can understand exactly how the AI arrives at its conclusions, fostering trust and facilitating regulatory approvals for clinical use. Transparency in AI decision-making processes is critical for medical applications, where diagnostic errors can have profound consequences.

The training of crossNN involved an extensive dataset encompassing the epigenetic profiles of over 8,000 reference tumors, each represented by hundreds of thousands of data points derived from diverse sequencing methods. The model was rigorously tested on more than 5,000 tumor samples, demonstrating robust performance even when analyzing incomplete epigenetic profiles or data generated using different techniques and varying quality.

An especially notable breakthrough lies in the model’s compatibility with minimally invasive liquid biopsies. In cases of brain tumors, cerebrospinal fluid—obtained through lumbar puncture rather than brain surgery—can provide sufficient genetic material for epigenetic fingerprinting. Using rapid nanopore sequencing, the researchers successfully analyzed these cerebrospinal fluid samples to deliver diagnoses without the need for risky surgical interventions. For example, a patient presenting with double vision was diagnosed accurately with a central nervous system lymphoma, enabling immediate commencement of targeted chemotherapy.

The development of this AI diagnostic tool responds to an urgent clinical need. Cancer medicine is evolving towards highly personalized treatments, often targeting specific molecular pathways unique to tumor subtypes. Precise, rapid tumor classification not only guides therapy selection but also opens the door to enrolment in clinical trials for rare tumors that might otherwise be misdiagnosed or overlooked. Thus, the crossNN model may accelerate the implementation of tailored cancer therapies, improving patient outcomes significantly.

Looking ahead, the research consortium plans to validate crossNN through clinical trials at all eight German Cancer Consortium (DKTK) centers nationwide. These studies will evaluate the model’s intraoperative applications, potentially transforming surgical oncology by providing real-time, accurate tumor classification during operations. The researchers emphasize the scalability and cost-effectiveness of this approach, positioning it as an accessible diagnostic tool in routine oncological care worldwide.

Beyond brain tumors, the model’s ability to classify a vast array of tumors from diverse organs underscores its versatility. By integrating data from various DNA methylation platforms and sequencing technologies, crossNN demonstrates powerful cross-platform generalizability. This advance addresses a longstanding challenge in computational oncology, where heterogeneous data sources often hampered the development of reliable AI models.

The study reflects a successful fusion of molecular biology, bioinformatics, and machine learning. Bioinformatician Dr. Sören Lukassen noted that while many prior AI models were complex and opaque, crossNN balances simplicity with high precision. This strategic design choice ensures broader acceptance in clinical settings, where explainability remains a major hurdle for implementing AI tools.

Additionally, the open-access crossNN user platform offers practitioners worldwide an opportunity to utilize the model for tumor classification, fostering collaborative advancements and feedback loops to refine its diagnostic power. The platform serves as an interface between cutting-edge computational research and frontline clinical practice, bridging the gap that often exists between laboratory innovations and patient care.

In conclusion, the crossNN AI framework marks a significant stride in oncological diagnostics, leveraging the epigenetic codes embedded in tumor DNA to deliver fast, accurate, and non-invasive tumor classification. Its explainability, robustness, and adaptability position it as an indispensable tool that could soon become integral to personalized cancer medicine, reshaping treatment pathways and offering hope for patients with previously challenging tumor diagnoses.

Subject of Research: Not applicable

Article Title: crossNN is an explainable framework for cross-platform DNA methylation-based classification of tumors

News Publication Date: 6-Jun-2025

Web References:

• Original publication – Nature Cancer
• Department of Neuropathology – Charité
• Interdisciplinary Tumor Boards – Charité Comprehensive Cancer Center
• BIH Medical Omics
• crossNN user platform

References:
Yuan D et al. crossNN is an explainable framework for cross-platform DNA methylation-based classification of tumors. Nature Cancer. 2025 June 06. doi: 10.1038/s43018-025-00976-5

Image Credits: © Charité | Philipp Euskirchen

Keywords: AI tumor classification, epigenetics, DNA methylation, crossNN, brain tumor diagnosis, liquid biopsy, nanopore sequencing, machine learning oncology, explainable AI, personalized cancer medicine

Colorectal cancer is one of the most common malignancies worldwide and remains a leading cause of cancer-related mortality. Its multifactorial etiology includes a complex interplay of genetic mutations alongside epigenetic alterations—heritable yet reversible changes in gene expression without modifications to the underlying DNA sequence. Among these mechanisms, DNA methylation at CpG islands within gene promoter regions plays a pivotal role by silencing tumor suppressor genes or regulatory non-coding RNAs, tipping the cellular environment toward malignancy.

The study meticulously evaluated the methylation status of the miR-139-5p promoter region in tumor tissues and corresponding plasma samples from patients diagnosed with CRC. miR-139-5p, known for its tumor-suppressive functions, was previously hypothesized to be epigenetically silenced in various cancers, but concrete data in colorectal cancer were lacking. Employing the highly sensitive MethyLight technique, which quantifies DNA methylation levels with precision, the researchers analyzed 80 paired samples of tumorous and adjacent healthy tissues, along with matched plasma specimens.

To assess the functional consequences of methylation, expression levels of miR-139-5p were measured via quantitative PCR (qPCR), providing a robust correlation between gene silencing and epigenetic modification. Furthermore, the study investigated the concentration of RAP-1b protein, a direct target gene suppressed by miR-139-5p, through ELISA assays, thereby linking molecular changes to phenotypic outcomes relevant in oncogenesis.

Results displayed a striking difference in methylation between cancerous and non-cancerous plasma samples, with median percentage of methylated reference (PMR) values approximately 12.4 in CRC patients versus as low as 0.66 in controls. This disparity underscores the robustness of miR-139-5p promoter methylation as a biomarker signature. Sensitivity and specificity for CRC detection in plasma samples were calculated at 75% and 92.5% respectively, yielding an outstanding area under the receiver operating characteristic curve (AUC) of 0.958, which highlights the excellent diagnostic potential of this epigenetic marker.

Additionally, the inverse relationship between miR-139-5p expression and RAP-1b protein concentration bolsters the molecular narrative where hypermethylation leads to silencing of the microRNA, thus relieving repression of oncogenic RAP-1b, contributing to tumor proliferation and progression. The statistically significant decrease in miR-139-5p expression in both plasma and tumoral tissue (<0.001, p-value) confirms the functional impact of methylation-mediated gene silencing within the CRC pathophysiological framework.

The researchers emphasize that plasma-based detection of miR-139-5p hypermethylation offers a minimally invasive and potentially cost-effective strategy for early CRC diagnosis. Current screening methods such as colonoscopy, although highly effective, face limitations including invasiveness, cost, and patient compliance. Liquid biopsy approaches, analyzing circulating tumor-derived DNA in blood, represent the frontier in precision oncology diagnostics. This study’s findings mark a critical stride toward integrating epigenetic biomarkers into routine clinical workflows.

While promising, the authors caution that additional large-scale prospective studies are essential to validate these findings across diverse populations and to evaluate longitudinal changes during treatment and disease remission. Determining the stability and dynamics of miR-139-5p methylation in plasma over time will be crucial to establishing its utility not only as a diagnostic tool but possibly as a prognostic or surveillance marker.

Moreover, this work aligns with increasing recognition of microRNAs’ central roles in cancer biology. These small non-coding RNAs regulate gene expression post-transcriptionally and are often dysregulated in malignancies. Epigenetic repression of microRNAs, such as miR-139-5p, illustrates a mechanism by which cancer cells circumvent tumor suppressive pathways, a concept that may extend to other miRNAs and cancer types.

Technically, the employment of MethyLight technology represents a gold standard for methylation quantitation, capable of detecting low-abundance methylated molecules even in heterogeneous samples like plasma. Coupled with qPCR and ELISA, the study offers a comprehensive, multimodal approach to dissecting the miR-139-5p/RAP-1b axis at DNA, RNA, and protein levels.

The integration of epigenetic biomarkers with conventional diagnostic methods could enhance sensitivity and specificity, reducing false positives and negatives that plague current cancer screening tests. Furthermore, since epigenetic changes are reversible, understanding the methylation landscape opens avenues for therapeutic interventions using demethylating agents, tailoring treatment to molecular profiles.

This study’s implications transcend colorectal cancer as the paradigm of epigenetic microRNA silencing could be applicable to multiple tumor entities where miR-139-5p or similar microRNAs function. Personalized medicine stands to benefit tremendously by incorporating such biomarkers into decision-making algorithms, optimizing patient outcomes.

In conclusion, the hypermethylation of miR-139-5p promoter DNA emerges as a highly promising plasma-based biomarker for colorectal cancer detection with impressive diagnostic accuracy. The work propels us closer to realizing non-invasive, precise blood tests capable of early cancer diagnosis, ultimately reducing mortality through timely intervention. The scientific community eagerly awaits further validation studies to transform this molecular insight into clinical reality, heralding a new era in oncologic diagnostics driven by epigenetic research.

Subject of Research: Epigenetic silencing of the microRNA gene miR-139-5p and its role in colorectal cancer pathogenesis and plasma-based diagnostic biomarker potential.

Article Title: Evaluation of epigenetic silencing of the miR-139-5p gene in the pathogenesis of colorectal cancer and its diagnostic biomarker capability in plasma samples

Article References:
Asefi, M., Rezvani, N., Saidijam, M. et al. Evaluation of epigenetic silencing of the miR-139-5p gene in the pathogenesis of colorectal cancer and its diagnostic biomarker capability in plasma samples. BMC Cancer 25, 877 (2025). https://doi.org/10.1186/s12885-025-14290-x

Image Credits: Scienmag.com

DOI: https://doi.org/10.1186/s12885-025-14290-x