The review, summarized in a Scientific Reports abstract, examined the incidence and metastatic rate of intramammary lymph nodes, their diagnostic performance, associated risk factors and prognostic findings. The pooled incidence of pathologically detected intramammary lymph nodes was 0.03 (95% confidence interval, 0.01–0.08), while the pooled metastatic rate was 0.32 (95% confidence interval, 0.27–0.37). The abstract did not specify the denominator underlying the reported metastatic rate, so it should not be interpreted as a rate for the entire study population.

For diagnostic performance, the meta-analysis reported pooled sensitivity of 0.505, specificity of 0.813, positive predictive value of 0.674 and negative predictive value of 0.711. It also reported a positive likelihood ratio of 2.738, a negative likelihood ratio of 0.635 and a diagnostic odds ratio of 4.961. These pooled indices describe the overall performance reported across the included studies rather than the performance in an individual patient. The authors concluded that the diagnostic accuracy of intramammary lymph node metastasis was insufficient to replace standard axillary staging.

Pooled analyses also identified lymphovascular invasion, tumor size or T stage, and histological grade as factors commonly associated with intramammary lymph node metastasis. Prognostic findings were inconsistent across the evidence, and the review reported that the independent prognostic value of intramammary lymph node metastasis remained inconclusive. The review therefore supports a strong association with axillary involvement but does not establish a uniform independent effect on patient outcomes. The available evidence is limited to the study-level results presented in the abstract, and the conclusion is based on 24 studies with a combined population of 1,482 patients.

Subject of Research: Systematic review of intramammary lymph node metastasis and axillary involvement in women with primary breast cancer

Article Title: Clinical significance of intramammary lymph nodes in female primary breast cancer: a systematic review and meta-analysis

Article References: Zhao, Z., Wang, X., Zhang, W., Zhang, L., Pei, S., Yang, Y., Qu, Z., Li, Z., Lu, J., Fang, Y., Wang, J., & Wang, Z. (2026). Clinical significance of intramammary lymph nodes in female primary breast cancer: a systematic review and meta-analysis. Scientific Reports. https://doi.org/10.1038/s41598-026-70289-9

Image Credits: AI Generated

DOI: 10.1038/s41598-026-70289-9

Keywords: primary breast cancer, intramammary lymph nodes, axillary metastasis, systematic review, meta-analysis, diagnostic accuracy, lymphovascular invasion, prognosis, Clinical, significance, intramammary, lymph

The investigators reviewed patients who had surgery at Shizuoka Cancer Center between September 2014 and September 2015. Eligible cases included stage I to stage IIIA non-small cell lung cancer, while patients receiving adjuvant platinum-based chemotherapy or immunotherapy were excluded. That restriction was intended to reduce the influence of postoperative systemic treatment on the assessment of immune variables. The final cohort was relatively small and centered on early-stage disease: 75.9% of tumors were pathological stage I and 75.2% were adenocarcinomas. Median follow-up for patients alive at their last visit was 60 months, with a reported range of 7 to 74 months. Disease-free survival was the primary endpoint, measured from surgery to recurrence or death, while overall survival was a secondary endpoint measured to death from any cause or the last follow-up visit.

To perform the digital analysis, the researchers selected formalin-fixed, paraffin-embedded surgical tissue blocks containing each tumor’s center and invasive margin. Slides were scanned at high resolution, and the investigators used HALO image-analysis software, whose tissue-classification system is based on a Random Forest machine-learning algorithm. The system used an ensemble of eight random forests, each containing 10 trees, with predictions averaged across the ensemble. Each forest used four default features: two color features based on Gaussian analysis and two texture features based on Hessian-of-Gaussian and structure-tensor analysis. Pathologists manually annotated tumor areas on hematoxylin-and-eosin-stained whole-slide images and excluded large necrotic regions where those areas could not be distinguished reliably from epithelium on AE1/AE3 pancytokeratin staining.

The same tissue classifier then separated epithelial, stromal, and non-epithelial/non-stromal regions by comparing the whole-slide images of the H&E and AE1/AE3-stained sections. The last category mainly included minor necrosis and mucus, and its lymphocytes were not included in the analysis. Separate CD8 and Foxp3 slides were processed with pathologist-selected parameters for nuclear segmentation and staining-intensity thresholds. The researchers then used the HALO multiplex immunohistochemistry algorithm to count CD8-positive and Foxp3-positive tumor-infiltrating lymphocytes separately in the epithelial and stromal compartments. Densities were reported as positive cells per square millimeter, rather than as counts from selected high-power fields. The PD-L1 tumor proportion score was calculated as the percentage of positive malignant cells among all malignant cells in the epithelial region.

Measurements were made across the whole annotated tumor region in one operation. The investigators also combined epithelial and stromal areas, but emphasized that this combined measurement did not represent the entire tumor because the amount of unevaluated necrosis differed among specimens. Across the combined regions, median CD8-positive density was 432.33 cells per square millimeter, compared with 335.59 cells per square millimeter in epithelium and 582.06 in stroma. Median Foxp3-positive density was 123.66 cells per square millimeter in the combined regions, with medians of 73.23 in epithelium and 179.20 in stroma. Median PD-L1 tumor proportion score was 16.27%, and values ranged from 0.093% to 77.45%. Epithelial and stromal CD8-positive densities were strongly correlated, with a Pearson coefficient of 0.81 and p<0.001; the corresponding Foxp3-positive correlation was 0.52, also with p<0.001. High CD8-positive density was associated with adenocarcinoma at p=0.011, while most other clinicopathological characteristics showed no significant relationship with the measured immune factors.

For survival comparisons, patients were divided at the cohort median for each immune measurement. Higher CD8-positive density was associated with better disease-free and overall survival whether the density was measured in epithelium, stroma, or both regions. For disease-free survival, the log-rank test yielded p<0.001 for the combined and epithelial CD8 measurements and p=0.003 for the stromal measurement. In multivariable Cox models, high combined CD8-positive density was independently associated with lower disease-free-survival hazard, with a hazard ratio of 0.43 and 95% confidence interval of 0.26 to 0.72. High PD-L1 tumor proportion score was also independently associated with disease-free survival, with a hazard ratio of 0.49 and 95% confidence interval of 0.30 to 0.81, as was high combined Foxp3-positive density, with a hazard ratio of 0.55 and 95% confidence interval of 0.34 to 0.89. Pathological stage remained an independent factor in each model; stage II or III disease had a disease-free-survival hazard ratio of 3.65 to 4.09 relative to stage I, depending on the immune variable in the model.

The overall-survival analysis produced a more selective pattern. High CD8-positive density remained independently associated with lower mortality hazard, with a hazard ratio of 0.37 and 95% confidence interval of 0.19 to 0.72, and high PD-L1 score did as well, with a hazard ratio of 0.48 and 95% confidence interval of 0.25 to 0.90. The association for combined Foxp3-positive density was not statistically significant in the multivariable analysis: its hazard ratio was 0.61, with a 95% confidence interval of 0.33 to 1.143 and p=0.11. Compartment-specific, unadjusted analyses showed that Foxp3-positive density in epithelium was associated with better disease-free and overall survival, whereas stromal Foxp3-positive density was not significantly associated with either endpoint. Thus, the location of that immune-cell measurement affected the observed results, while the study’s multivariable analysis did not establish Foxp3-positive density as an independent overall-survival marker.

Stage-specific analyses indicated that the digital immune measurements could distinguish prognosis among patients whose cancers shared the same broad TNM category. Among stage I patients, five-year disease-free survival was 81.5% for those with high combined CD8-positive density and 57.6% for those with low density, a difference with p=0.003. The corresponding figures for high and low combined Foxp3-positive density were 81.3% and 59.1%, with p=0.035. Stage I patients with high PD-L1 score had a five-year disease-free survival of 80.8%, compared with 60.3% among those with a low score, although that difference did not cross the study’s statistical-significance threshold of p<0.05, with p=0.053. For overall survival in stage I disease, the high-versus-low five-year estimates were 90.9% versus 70.8% for CD8-positive density and 90.3% versus 72.9% for PD-L1 score. In multivariable analysis limited to stage I patients, high CD8-positive density was independently associated with disease-free survival, with a hazard ratio of 0.39 and 95% confidence interval of 0.20 to 0.77.

The findings come from a retrospective, single-institution cohort and have several limitations. The sample was relatively small, and the analysis did not examine pre-existing lung conditions such as emphysema or interstitial lung disease in detail. It evaluated only three features in a tumor microenvironment composed of many cell types. Bias may also have affected the pathologist’s development of the algorithm, and the manual exclusion of substantial necrosis meant that the epithelial and stromal areas did not equal the whole tumor. The use of median-derived cutoffs limits what can be inferred about specific thresholds, while the selected Japanese population and exclusion of patients receiving adjuvant treatment restrict how broadly the results can be applied. The study therefore establishes associations and the feasibility of automated immune quantification in surgical specimens; it does not demonstrate that the measurements caused better outcomes or determine which postoperative treatment would benefit an individual patient.

Despite those constraints, the work shows how computational pathology can make spatial immune measurements across a resection specimen more systematic than manual counting in selected microscopic fields. CD8-positive density had the most consistent relationship with outcomes across the analyses, while epithelial and stromal CD8 measurements were strongly correlated and each was associated with survival. Measuring the two compartments together also retained prognostic associations and could simplify analysis, although the study did not establish that this approach provided a statistically stronger separation than separate assessment. The authors noted that more advanced deep-learning systems might distinguish tumor and nontumor regions automatically and reduce manual annotation. For the present evidence, the central result is that machine-learning-based whole-slide analysis identified reproducible, spatially resolved immune measures that supplied prognostic information beyond pathological stage, while their role in postoperative treatment selection remains to be established.

Subject of Research: Retrospective prognostic analysis of digitally measured tumor immune features in surgically resected non-small cell lung cancer

Article Title: Prognostic significance of tumor microenvironment assessed by machine learning algorithm in surgically resected non‐small cell lung cancer

Article References: Terada, Y., Isaka, M., Ono, A., Kawata, T., Serizawa, M., Mori, K., Muramatsu, K., Tone, K., Kenmotsu, H., Ohshima, K., Urakami, K., Nagashima, T., Kusuhara, M., Akiyama, Y., Sugino, T., Takahashi, T., & Ohde, Y. (2026). Prognostic significance of tumor microenvironment assessed by machine learning algorithm in surgically resected non‐small cell lung cancer. Cancer Reports, 9(10), Article e1926. https://doi.org/10.1002/cnr2.1926

Image Credits: AI Generated

DOI: 10.1002/cnr2.1926

Keywords: non-small cell lung cancer, machine learning, digital pathology, whole-slide imaging, tumor-infiltrating lymphocytes, CD8-positive T cells, Foxp3-positive T cells, PD-L1, cancer prognosis, computational pathology, Prognostic, significance

The specimen was captured in July 2025 during a series of chiropteran surveys in southwestern Yunnan. Sweep netting at the eaves of the management station building within the Jinguang Temple Provincial Nature Reserve in Dali, at coordinates 25°11’55″N, 99°31’50″E and an elevation of roughly 2,500 metres, yielded one adult female, catalogued as GZHU 25209. Field protocols complied with Chinese law and live-animal handling followed American Society of Mammalogists guidelines. The voucher specimen, preserved in 75 percent ethanol, is now deposited at the South China Biodiversity Research Center at Guangzhou University. Adult status was confirmed by the degree of epiphyseal-diaphyseal fusion in the bones, a standard ageing method in bats.

Identifying the animal required an integrative approach combining morphology, morphometrics and genetics. The team took external measurements including body length, tail length, ear length, hind foot length, forearm length and tibia length, plus twelve craniodental dimensions measured with digital calipers accurate to 0.01 millimetres. The right wing was pinned flat, photographed perpendicular to the plane, and analysed with BAT-WINGAREA software to derive total wing area, wingspan and membrane areas, from which the three primary airfoil parameters were calculated. For the genetic component, total DNA was extracted and the mitochondrial cytochrome b gene, a 1,140-base-pair fragment widely used in bat systematics, was amplified and sequenced with specific primers.

The phylogenetic analysis aligned the new sequence with publicly available data from related genera including Pipistrellus, Vespertilio, Nyctalus and Falsistrellus, using Myotis chinensis as an outgroup. Maximum likelihood inference in IQ-TREE, with a TIM2+F+I substitution model and ultrafast bootstrap support estimated over 1,000 replicates, placed the Yunnan specimen firmly within a clade containing published sequences of Peters’s Pipistrelle from Luzon Island in the Philippines, with a bootstrap value of 100, the maximum possible. Pairwise genetic distances under the K2P model showed the new sample differed from Philippine H. petersi by only 0.10, whereas distances to other Hypsugo species ranged from 0.12 to 0.29. Notably, the Chinese sample formed the basal lineage relative to other conspecifics, hinting at an older, possibly ancestral population in the region.

Morphology told a consistent story. The bat is large for the genus, with a forearm length of 39.94 millimetres and a body mass of 7.5 grams. Its dorsal fur is long and woolly, dark chocolate-brown with lighter tips that give a frosted appearance, while the ventral fur is paler with brown tips. The wing membranes are blackish and hairless except for a small patch of pale-brown fur extending onto the ventral surface of the femur and the area around the anus, a hallmark of the species. The ears are dark, broad and rounded, the tragus short, broad and blunt, and the calcar bears a slight keel. The skull is long and narrow with slender zygomatic arches, well-developed occipital crests, and lower molars of the myotodont type, in which the talonid exceeds the trigonid in size, a diagnostic feature of Hypsugo that distinguishes it from Pipistrellus.

Multivariate statistics added a third line of evidence. Principal component analysis of six external measurements, supplemented with data from four other Chinese Hypsugo species, explained 69.6 percent of total variance in the first two components and separated five morphological groups, with the new specimen distinguished chiefly by its relatively long ears. A second PCA based on twelve craniodental measurements explained 72.1 percent of variance and clearly separated the Yunnan bat from all known Chinese Hypsugo, with condylobasal length and width across the upper canines driving the separation. Some measurements were slightly larger than previously published ranges, but the combined morphological, molecular and multivariate evidence left little doubt about the identification.

The taxonomic history of this bat is convoluted, which partly explains why it has been so easy to overlook. Described as Vespertilio petersi in 1899 from northern Sulawesi, the species was shuffled through the twentieth century between Pipistrellus and the genus Falsistrellus, erected in 1943 for Australian relatives. Hill and Harrison in 1987 even suggested that Asian Falsistrellus might be the eastern representative of Hypsugo. Only in 2018 did a comprehensive study integrating molecular, cranial, dental and multivariate evidence formally transfer the Asian members of Falsistrellus, including petersi, into Hypsugo, a revision now widely adopted. The genus Hypsugo itself comprises 22 species distributed across the southern Palaearctic, most of Africa and the Indomalayan region.

Because confirmed records of H. petersi remain so few, the International Union for Conservation of Nature classifies the species as Data Deficient, with critical gaps in knowledge about population trends, ecology and threats. The new record comes from southern Yunnan, adjacent to the Indo-China Peninsula biodiversity hotspot spanning Laos, Myanmar and Vietnam. The complete absence of records across that vast peninsula suggests a significant distributional gap and indicates that the species’ true geographic range is far less understood than its scattered island records imply. Targeted surveys and molecular screening of museum specimens and field samples across the region are now needed to map where this bat actually occurs.

Wing morphology offers clues to how the species lives. With high wing loading of about 16.77 newtons per square metre, an average aspect ratio of about 6.87 and a low tip-shape index, H. petersi is built for moderately fast flight with limited maneuverability, suited to foraging along semi-open corridors such as forest edges and gaps. This matches the field observation of the individual roosting in a building crevice surrounded by tall, dense coniferous forest at 2,539 metres elevation, and it aligns with earlier accounts of the species being caught in mossy and pine forests at high elevations in the Philippines and roosting under house eaves in Borneo and the Philippines. Remarkably, the Yunnan record extends the known altitudinal limit of the species upward by 400 metres.

To help future surveys, the researchers also published the first comprehensive identification key to all seven Hypsugo species now known from China, in both English and Chinese. Previous guides covered only three species, and the transfer of Asian Falsistrellus species to Hypsugo in 2018 had left the Chinese fauna without an up-to-date diagnostic tool. The key relies on forearm length, fur coloration on the femur and perianal region, dental characters such as the position of the first upper premolar, and tragus shape. The authors caution that the presence of two large-bodied relatives, H. affinis and H. mordax, remains unverified in China and flag them as requiring confirmation. The discovery underscores that bat diversity in China is still incompletely documented and that integrative taxonomy, combining genetics, morphology and morphometrics, remains essential for uncovering it.

Subject of Research: First record of the bat Hypsugo petersi in China based on morphological and genetic evidence

Article Title: First record of the Peters’s Pipistrelle Hypsugo petersi ( Chiroptera , Vespertilionidae ) from China and an updated identification key to Chinese Hypsugo

Article References: Guo, Y., Chen, K.-H., Fan, X.-D., Xue, L.-Y., Yang, H.-H., Xiao, Y.-J., Wang, X.-Y., Wu, Y., Yue, Y., & Yu, W.-H. (2026). First record of the Peters’s Pipistrelle Hypsugo petersi (Chiroptera, Vespertilionidae) from China and an updated identification key to Chinese Hypsugo. ZooKeys, 1290, 171-184. https://doi.org/10.3897/zookeys.1290.186733

Image Credits: AI Generated

DOI: 10.3897/zookeys.1290.186733

Keywords: Hypsugo petersi, bats, Chiroptera, Yunnan, China, new record, cytochrome b, phylogenetics, morphometrics, conservation, ZooKeys, taxonomy

Routine inflammatory markers used to gauge pneumonia severity have limited specificity for the lung damage caused by C. pneumoniae. White blood cell counts, high-sensitivity C-reactive protein, and procalcitonin are commonly measured, but they reflect systemic inflammation rather than local alveolar epithelial injury. Because C. pneumoniae triggers a chronic, monocyte-macrophage-predominant non-necrotic inflammation localized mainly to the airway mucosa, many of these markers remain normal even when the infection is severe. A team of clinical laboratory researchers in Suzhou, China, led by Lei Ding of the Children’s Hospital of Wujiang District and collaborating with colleagues at The Fifth People’s Hospital of Suzhou, Nanjing Medical University, and Suzhou Municipal Hospital, set out to determine whether a different kind of biomarker could fill this gap.

The marker they investigated is Krebs von den Lungen-6, or KL-6, a large molecular weight glycoprotein antigen mainly secreted by proliferating or damaged type II alveolar epithelial cells. KL-6 corresponds to the shed ectodomain of mucin 1, a transmembrane glycoprotein constitutively expressed on the apical surface of type II alveolar epithelial cells. Under both physiological and pathological conditions, circulating KL-6 concentrations are determined by basal and stress-induced MUC1 synthesis coupled with tightly regulated ectodomain shedding. During pulmonary inflammation or fibrotic remodeling, matrix metalloproteinase-mediated proteolytic cleavage of MUC1 releases the soluble KL-6 ectodomain into the alveolar space, and subsequent spillover into the systemic circulation produces measurable elevations in serum levels. The Japanese scholar Kohno and colleagues first used KL-6 in 1989 to evaluate disease activity in interstitial lung disease, and in 1999 the Japanese National Health Insurance System officially included it as a biomarker for assessing ILD disease activity. Since then, KL-6 has been shown to predict the risk of severe COVID-19 pneumonia, correlate with in-hospital mortality in community-acquired pneumonia, and monitor disease progression in immune checkpoint inhibitor-induced pneumonitis. Its expression pattern in acute C. pneumoniae infection in children, however, remained unclear.

To address this, the researchers enrolled 68 hospitalized children with acute lower respiratory tract infection caused by C. pneumoniae at the Children’s Hospital of Wujiang District between May 2024 and April 2025. The cohort comprised 48 mild cases and 20 severe cases, with 19 healthy children serving as controls. Inclusion required a clinical diagnosis based on the Zhu Futang practice of pediatrics, positive serum C. pneumoniae-IgM, and imaging findings of segmental pneumonia or extensive unilateral or bilateral lesions. Severe cases were defined by poor general condition, refusal to eat or dehydration, consciousness disturbance, rapid respiratory rate, cyanosis, dyspnea, pulmonary infiltration involving multiple lobes or at least two-thirds of the lung, pleural effusion, pulse oxygen saturation at or below 0.92, or extrapulmonary complications. Children with allergies, tumors, immunodeficiency, or other congenital diseases were excluded, as were those with mixed infections involving Mycoplasma pneumoniae, influenza A or B virus, respiratory syncytial virus, human rhinovirus, or adenovirus, confirmed by multiplex PCR on throat swabs.

Serum KL-6 levels were measured using a Fujifilm LUMIPULSE G1200 chemiluminescent immunoassay system with a lower limit of detection of 1 U/mL and an intra-assay coefficient of variation of at most 4.4 percent. C. pneumoniae-IgM was quantified on an iFlash 3000 chemiluminescent immunoassay system, and routine blood tests, biochemical parameters, coagulation, and infection markers were performed on automated analyzers from Mindray, Sysmex, and Roche. Imaging examinations used a GE Optima 640 whole-body CT scanner, and the number of infected lung segments was collected retrospectively. Statistical analyses were carried out with GraphPad Prism 8.0 and IBM SPSS Statistics 22, with normally distributed data compared using the t test and non-normally distributed data analyzed with the Wilcoxon rank-sum test. Diagnostic performance was assessed using ROC curves, and the 95 percent confidence interval of the area under the curve was calculated with the Percentile Bootstrap method, generating 1000 ROC curves through repeated sampling with replacement to avoid the limitation of the traditional normal approximation when the AUC is close to 1 and the sample size is small.

The results showed that serum KL-6 levels were significantly higher in the severe group than in the mild group, with a median of 303.00 U/mL in severe cases compared with 146.50 U/mL in mild cases, a difference that was highly statistically significant. Binary logistic regression analysis, which entered a maximum of two predictor variables following the ten events per variable rule to mitigate overfitting risk, indicated that KL-6 was independently associated with severe C. pneumoniae infection in children, with an odds ratio of 1.048 per unit increase. The area under the ROC curve for differentiating severe from mild infection was 0.951, with a 95 percent confidence interval of 0.880 to 0.997, indicating a statistically significant capacity to discriminate disease severity. The optimal cut-off value was 196.5 U/mL, which fell below the proposed pediatric benchmarks of 250 U/mL reported by Imai and colleagues and 300 U/mL proposed by Kubota and Haruta.

Further analysis revealed that the KL-6 level was positively correlated with the number of pulmonary lesions, the number of hospitalization days, and C. pneumoniae-IgM levels, but not with other indicators such as CRP, PCT, or IL-6. Children with fewer than three lesions had significantly lower KL-6 levels than those with three or more lesions. When the cohort was split into high and low expression groups based on the median KL-6 level of 159 U/mL, the high KL-6 group exhibited longer hospitalization duration and higher C. pneumoniae-IgM levels, whereas no inter-group differences were observed for other laboratory indicators. This pattern suggests that KL-6 mainly reflects pulmonary parenchymal injury rather than the systemic inflammatory response, compensating for the deficiency of routine indicators in evaluating local lung injury.

Interestingly, serum lactate dehydrogenase was marginally lower in the severe group than in the mild group, but both values fell within the healthy pediatric reference range, and the inter-group difference was smaller than the intrinsic biological variation among healthy children. The researchers attribute this borderline statistical result partly to the small sample size of severe cases and retrospective sampling on admission. The finding fits the pathogenic features of C. pneumoniae, which triggers chronic, non-necrotic inflammation localized mainly to the airway mucosa without widespread acute pulmonary parenchymal necrosis, thereby limiting systemic LDH release. Multiple pediatric cohorts corroborate this view, with most inflammatory markers normal in 145 children with C. pneumoniae pneumonia, and in a 141-patient risk model only wheezing and co-infection were independent predictors of severe disease, whereas necrosis-associated markers such as LDH were excluded.

The study is the first to confirm that serum KL-6 is independently associated with severe acute C. pneumoniae infection in children, and the results are consistent with the role of KL-6 in other lung diseases, including Mycoplasma pneumoniae pneumonia and respiratory syncytial virus bronchiolitis. The average KL-6 expression level in patients with C. pneumoniae infection is lower than that in patients with Mycoplasma pneumoniae pneumonia, possibly because C. pneumoniae infection causes mainly mild to moderate lung injury and the study excluded children with mixed infections, reducing the inflammatory effect. The authors suggest that if KL-6 levels continue to increase, severe infection should be suspected and timely intervention implemented. KL-6 does not need to be a routine screening indicator given that C. pneumoniae infection is mainly mild, but for children with severe symptoms such as persistent cough and rapid breathing, KL-6 detection can improve the accuracy of disease assessment and optimize the allocation of medical resources.

The researchers acknowledge several limitations. The study was single-center and small, with only 20 severe cases among 68 enrolled children, which may have affected statistical power. Changes in KL-6 levels before and after treatment were not tracked, so the value of KL-6 in monitoring treatment effects could not be evaluated. Children with mixed infections or underlying diseases were excluded, so the results may not generalize to complex cases, and no mechanistic experiments were performed to determine how KL-6 is released during alveolar epithelial cell injury, meaning the findings establish associations rather than causal relationships. Future work, the authors suggest, should include multicenter prospective investigations, longitudinal serial monitoring of KL-6 dynamics, inclusion of patients with complex comorbidities or mixed infections, and rigorous mechanistic dissection of MUC1-mediated KL-6 release pathways. Published in Frontiers in Cellular and Infection Microbiology on 16 October 2026, the study provides a basis for the rational selection of clinical detection items and positions KL-6 as an auxiliary biomarker for assessing pulmonary parenchymal injury in suspected severe C. pneumoniae cases.

Subject of Research: Biomarkers

Article Title: Clinical significance and discriminatory performance of the serum KL-6 level as a biomarker for severe Chlamydia pneumoniae infection in children

Article References: Ding, L., Chen, S., Song, H., Cheng, S., Shao, M., Zhang, H., Hua, J., Gao, F., & Xu, J. (2026). Clinical significance and discriminatory performance of the serum KL-6 level as a biomarker for severe Chlamydia pneumoniae infection in children. Frontiers in Cellular and Infection Microbiology, 16, Article 1791097. https://doi.org/10.3389/fcimb.2026.1791097

Image Credits: AI Generated

DOI: 10.3389/fcimb.2026.1791097

Keywords: Clinical, significance, discriminatory, performance, serum, KL-6, level, biomarker, severe, Chlamydia, pneumoniae, infection

The research team, led by Thomas E. White and Noah Brown, focused on three taxa that co-occur across coastal and montane New South Wales: Hadronyche cerberea, Hadronyche versuta and members of the Atrax robustus complex. The latter grouping is treated at the complex level because Atrax robustus was recently split into three species, A. montanus, A. robustus and A. christenseni, and webs cannot yet be reliably assigned beyond the complex in the field. Together, the three focal taxa span the major architectural categories previously described for the family, including entrance-associated, lattice and open-hole web forms, making them a tractable comparative system that captures much of the recorded structural diversity in Atracidae.

Methodologically, the study is notable for what it did not do. Consistent with the medical risk posed by atracids, the researchers never collected, handled or excavated a single spider. Instead, they systematically searched eight sites across three regions of New South Wales, the Central Coast, Greater Sydney and the Southern Highlands, photographing burrows in embankments, logs, crevices and tree cavities with a smartphone camera held as close to perpendicular to the web surface as possible, with a scale reference in every frame. Species identity was assessed from web and burrow architecture combined with the external morphology of spiders visible at burrow entrances, evaluated against regional taxonomic keys. Image screening removed photographs with substantial obscurity, poor focus or incomplete structures, leaving final samples of 13 webs for the Atrax robustus complex, 13 for H. cerberea and 9 for H. versuta.

The analytical workflow converted those photographs into numbers. Each calibrated image was imported into Fiji, an open-source image analysis platform, and the researchers delineated two functionally distinct regions of interest. The sensory zone was defined as the full extent of the web, including all radiating triplines, representing the area over which the spider could plausibly detect substrate-borne vibrations from prey. The capture zone was defined as the subset of silk near the burrow entrance showing visibly denser threads, corresponding to the region most likely involved in physically intercepting prey. This division maps directly onto the two phases of atracid hunting: these are sit-and-wait ambush predators that remain inside their burrows until prey contact a tripline or entrance silk, at which point they rush out, seize the prey with their chelicerae and drag it back underground.

From convex hull polygons traced around each zone, the team extracted four morphometric metrics: area, circularity, aspect ratio and the sensory-to-capture allocation ratio. Principal component analysis showed that the first two components explained 68.6 percent of total variation, with the sensory-to-capture ratio dominating the first axis and overall web area and shape dominating the second. Permutational multivariate analysis of variance confirmed statistically significant differences both between web zones and among species, with all pairwise comparisons among the three taxa significant. The numbers affirm what field naturalists had long suggested qualitatively: these species build genuinely different webs, and the differences are consistent enough to measure.

The species-level patterns are striking. Hadronyche cerberea, found exclusively in arboreal contexts on the Central Coast, where it occupies fissures, heartwood pipes and borer holes in trees such as Casuarina, Melaleuca, Banksia and Eucalyptus, builds the largest webs of the three, with a mean sensory zone of 13,230 square millimetres and a mean capture zone of 1,436 square millimetres. Its webs typically feature two to three entrances connected by a thick silk tube, each surrounded by a dense lattice that radiates outward in a fan-like arrangement, with long triplines extending in a wide-angled distribution. Its mean sensory-to-capture ratio of 23.94 was by far the largest and most variable of the three taxa, indicating heavy proportional investment in detection rather than interception.

The Atrax robustus complex, by contrast, builds the smallest and most irregular webs, with a mean sensory zone of 5,940 square millimetres, a mean capture zone of 870 square millimetres and the lowest sensory-to-capture ratio at 7.18. Its stereotypical design is asymmetrical, commonly with two to six entrances interconnected by a dense-mesh silk atrium, and concave crescent flaps held slightly open even when not in use. Triplines extend outward erratically, branching and intersecting, and lattice density rarely decreases with distance from the entrance. Hadronyche versuta occupies an intermediate position, with a relatively symmetrical web built around an open-hole entrance, a dense lattice of thick silk radiating fully around the burrow, and the highest circularity values in both zones, indicating the most regular outlines. Its sensory-to-capture ratio of 8.98 sits close to that of Atrax.

These architectural differences appear to encode different trade-offs between prey detection, capture efficiency and predator avoidance, although the authors are careful to note that their functional interpretations remain inferential, since they did not observe prey capture, quantify prey availability or manipulate web structure. The arboreal H. cerberea, facing reduced prey abundance and elevated predation risk from visual predators such as birds, appears to favour extensive sensory coverage over a large capture area, with a smaller capture zone limiting the distance the spider must travel from its camouflaged, reinforced silk tube. The Atrax complex invests proportionally more in its capture zone through fine silk sheets but has a reduced sensory range, while H. versuta, which lacks an obvious defensive structure despite facing comparable terrestrial predators, may compensate through earlier threat detection via a broadly radial distribution of triplines providing near-uniform angular coverage around its entrance.

Intraspecific variation was also evident, and along different axes for different species. H. cerberea varied primarily in functional investment, with individuals differing in capture zone allocation while maintaining a consistent overall shape, largely reflecting differences in silk tube configuration and entrance number. The Atrax complex varied most in web shape, showing plasticity in overall form while consistently emphasising the capture zone. H. versuta showed the least variation overall. Ontogeny appears to contribute as well: in both Hadronyche species, smaller and likely juvenile individuals built simpler, smaller designs with less developed lattices, a pattern consistent with observations in tarantulas where web complexity increases with maturity. Microhabitat also shaped structure, with tripline number and length in H. versuta depending on adjacent substrates around the burrow entrance.

Beyond its evolutionary interest, the study has immediate applied value. Burrow and web architecture has already proven a reliable diagnostic feature for distinguishing trapdoor spider genera such as Euoplos and Cryptoforis, including sympatric populations, and the consistent differences quantified here provide a similar basis for field identification of H. versuta, H. cerberea and the Atrax robustus complex without excavating individuals, a process that poses serious safety risks with atracids. Potential applications span ecological surveys, medical identification, antivenom collection and future research. The authors propose that expanding this work to include additional species, particularly the genus Illawarra, and integrating behavioural assays of prey detection and capture with architectural data would allow the functional hypotheses to be tested directly. Data and code are openly archived via Zenodo, and the quantitative framework is readily applicable across the Mygalomorphae, offering a path toward understanding how silk use has shaped the ecological and evolutionary success of one of the oldest lineages of spiders.

Subject of Research: Ecology

Article Title: Inter‐ and Intraspecific Variation in Web Architecture and Its Significance in Australian Funnel‐Web Spiders (Atracidae)

Article References: Brown, N., & White, T. E. (2026). Inter‐ and Intraspecific Variation in Web Architecture and Its Significance in Australian Funnel‐Web Spiders (Atracidae). Ecology and Evolution, 16(10), Article e74473. https://doi.org/10.1002/ece3.74473

Image Credits: AI Generated

DOI: 10.1002/ece3.74473

Keywords: Inter, Intraspecific, Variation, Architecture, Significance, Australian, Funnel, Spiders, Atracidae, scientific research

The key to the finding lies in a zone of the glacier that is only a few kilometers wide but exerts an outsized influence on its fate: the grounding zone. This is the region where the ice sheet, resting on the seafloor, transitions to floating ice. Traditional models treat the grounding line as a sharp, discrete boundary between grounded and floating ice. Yet recent high-resolution satellite observations from the US-Finnish ICEYE synthetic aperture radar constellation have revealed something different in reality. As tides rise and fall, the grounding line migrates back and forth, and seawater of warm, modified Circumpolar Deep Water origin penetrates several kilometers beneath what is nominally grounded ice, episodically lifting it and actively melting it from below.

To test whether this dynamic grounding zone matters for the glacier’s long-term evolution, the researchers used the Ice-sheet and Sea-level System Model (ISSM), a high-resolution model in which the mesh resolves features as small as 500 meters near the grounding line. They implemented an actively melting grounding zone in which basal melt decreases linearly from a maximum at the grounding line to zero at a specified intrusion distance, building on earlier numerical frameworks. They then ran an ensemble of 105 simulations spanning melt rates from 10 to 150 meters per year and grounding-zone lengths from zero to six kilometers, and asked a simple question: which combination of parameters can reproduce nearly three decades of observed retreat, thinning, and acceleration of Thwaites Glacier between 1996 and 2023?

The answer was unambiguous. Simulations without a grounding zone, or with an intrusion of only about one kilometer, systematically failed to match the observational record. They underestimated the glacier’s ice discharge and thinning between 2004 and 2022, missed the observed acceleration, and required implausibly high submarine melt rates of around 150 meters per year to even approximate the early retreat. In contrast, simulations with grounding zones two to five kilometers wide, combined with melt rates of roughly 90 to 120 meters per year — values consistent with independent satellite estimates and high-resolution ocean modeling — successfully reproduced the concurrent retreat of both the fast-flowing Thwaites Main Trunk and the slower Thwaites Eastern Ice Shelf, along with the roughly 32 meters of grounded-ice thinning observed over the period.

The team formalized this comparison using a Bayesian calibration framework, drawing 10,000 random samples of melt rate and grounding-zone length and weighting them by how well each reproduced four independent observables: grounding-line retreat, ice discharge, thickness change, and glacier speed. The resulting posterior distribution revealed a well-constrained region of parameter space centered on grounding-zone lengths of two to five kilometers and melt rates of 90 to 120 meters per year. Strikingly, simulations lacking a grounding zone were four times less likely to reproduce the historical trend than simulations with a kilometer-scale intrusion. The analysis also exposed a trade-off between the two parameters — longer intrusions can compensate for somewhat lower melt rates — underscoring the value of calibrating against multiple variables simultaneously rather than a single metric.

With the model validated against the historical record, the researchers extended the simulations to 2100 under a deliberately conservative, lower-bound forcing scenario: ocean and atmospheric conditions held constant at their 1996–2023 averages. Even under this no-extra-warming assumption, the projections diverged sharply from conventional results, and the reason traces back to the seafloor topography beneath the glacier. Thwaites’ main trunk currently rests against the Mouginot Ridge, a bathymetric high a few kilometers across, lying roughly 950 meters below sea level, that acts as a pinning point slowing the retreat. The simulations show that whether the glacier detaches from this stabilizing ridge is governed almost entirely by the length of the grounding zone.

Simulations without a grounding zone, or with a one-kilometer intrusion, never retreat beyond the Mouginot Ridge by 2100, regardless of the melt rate imposed. But simulations with two- to three-kilometer grounding zones show about a 45 percent probability of unpinning by century’s end, with average unpinning dates of 2088 and 2065 respectively. For grounding-zone lengths of four to six kilometers — consistent with the most recent satellite-inferred extents — the probability of crossing the ridge rises to roughly 60 percent, and at five to six kilometers the glacier can unpin as early as 2038. Once free of the ridge, the grounding line retreats across predominantly retrograde, deepening seafloor toward the next major topographic barrier, the South Ridge, with no indication of slowdown or stabilizing feedback anywhere in the ensemble.

The sea-level consequences follow directly. With no grounding zone represented, the model yields a contribution of just 8.3 ± 3.0 millimeters by 2100, below the mean of the ISMIP6 intercomparison ensemble. A narrow one-kilometer grounding zone brings the estimate to 14.3 ± 3.1 millimeters, comparable to the ISMIP6 mean of 15.2 ± 21.8 millimeters. But a three-kilometer grounding zone — the maximum of the Bayesian posterior — produces 19.4 ± 5.2 millimeters, a 130 percent increase over the ISMIP6 mean, and a five-kilometer zone, matching the latest observations, yields 30.1 ± 5.7 millimeters, twice the multi-model average. Sensitivity tests revealed that the projections are far more sensitive to grounding-zone extent than to ocean temperature: each additional kilometer of intrusion changes the projected sea-level contribution by about 22 percent on average, whereas a plausible 0.5 to 0.7 degree Celsius of future ocean warming would raise ice loss by only about 15 percent. Simply introducing a grounding zone at all increases the projected contribution by 51 percent relative to traditional formulations.

The study is not without caveats, which the authors lay out carefully. The grounding-zone length is held fixed in time, whereas in reality the intrusion extent likely varies with bed slope, tidal amplitude, and subglacial water pressure; the team interprets it as an effective, time-averaged scale that captures decadal behavior. The ice front is fixed, so calving and the possible loss of pinning points that buttress the floating ice shelf are excluded by design, allowing the isolated effect of grounding-zone melting to be quantified. The melt parameterization is simplified rather than a fully coupled ocean model, and basal friction is not reduced by seawater lubrication, which may lead the model to underestimate thinning and speed-up. Bathymetric uncertainties also required a localized correction of up to 100 meters to the seafloor elevation beneath floating ice, within the error bounds of gravity inversion methods, to achieve consistency with the observed 1996 grounding line.

Even so, the implications for sea-level science are difficult to overstate. The work demonstrates that a process operating over just a few kilometers — invisible to coarse models and long treated as a detail — fundamentally changes both how well models can reproduce the past and how much ice loss they project for the future. The authors argue that actively melting grounding zones should be systematically incorporated into regional and continental-scale ice-sheet models, including international intercomparison exercises such as ISMIP6, and that expanded in-situ measurements near grounding zones are urgently needed to constrain heat delivery to the ice-shelf cavities. For a glacier holding the equivalent of 65 centimeters of global sea-level rise, the message is clear: the most important action at Thwaites may be happening in the narrow, tidally flooded strip of seafloor where ocean and ice meet, and until models capture it, the world’s best projections of Antarctic sea-level contribution may be undercounting the risk.

Subject of Research: The role of kilometer-scale seawater intrusions at the grounding zone of Thwaites Glacier in driving ice retreat and sea-level rise

Article Title: Sea-level contribution from Thwaites Glacier doubles in an observation-constrained ice-sheet model

Article References: Poinelli, M., Rignot, E., Seroussi, H., Cuzzone, J., Pelle, T., & Larour, E. (2026). Sea-level contribution from Thwaites Glacier doubles in an observation-constrained ice-sheet model. Nature Communications, 17(1), Article 10364. https://doi.org/10.1038/s41467-026-77976-1

Image Credits: AI Generated

DOI: 10.1038/s41467-026-77976-1

Keywords: Thwaites Glacier, sea-level rise, grounding zone, Antarctica, ice-sheet modeling, seawater intrusion, ISSM, Bayesian calibration, Mouginot Ridge, ocean-driven melting, West Antarctic Ice Sheet, ISMIP6

The research team turned to lake sediments preserved within Croajingolong National Park, one of the most repeatedly burnt landscapes in southeastern Australia, to read the region’s environmental past. Lake sediments accumulate in annual and seasonal layers, trapping charcoal particles, pollen grains and eroded soil that together form a continuous archive of what grew on the land, how hot fires burned and how severely the ground eroded across centuries. By extracting cores from the lakebed and analysing these preserved signals layer by layer, the scientists reconstructed the past 1000 years of ecological change with enough resolution to distinguish the pre-colonial period, when Gunaikurnai people actively managed the landscape, from the post-colonial era, when that management ceased. The technique allowed the researchers to separate the influence of climate variability from the influence of land use, a distinction that shorter records spanning only the industrial era cannot reliably make.

What the sediment record revealed was a cascading sequence of changes that began with fire itself. Infrared spectroscopy of charcoal fragments preserved in the sediments provided a proxy for past fire temperatures, distinguishing the cool, low-intensity burns characteristic of cultural fire practice from the hot wildfires that followed. The record shows that after colonisation, fire temperatures rose sharply. Those hotter fires came first in the sequence, and they drove a four-fold surge in eucalypt cover, transforming what had been maintained as open woodland for centuries into dense, closed forest. As the canopy thickened and fuel accumulated on the forest floor, soil erosion in the catchment increased, providing independent physical evidence that the landscape had crossed into a fundamentally more disturbed state.

Critically, the researchers demonstrated that this ecological transformation predates human-driven climate change. Dangerous fire weather is not new to East Gippsland: climate records show that seasons comparable to those that produced the Ash Wednesday fires of 1983 and the Black Saturday fires of 2009 occurred repeatedly in the region before colonisation. Yet during those earlier episodes of severe fire weather, the open woodland maintained by Gunaikurnai cultural burning did not produce catastrophic fire. The fuel loads were simply too low. It was only after colonisation ended that active management, allowing forests to expand and thicken across the region, that the same dangerous weather conditions began translating into landscape-scale disaster. In the authors’ framing, colonisation loaded the landscape with fuel, and it is that loaded landscape that has handed control of fire to the climate.

Lead author Professor Michael-Shawn Fletcher, of the School of Geography, Earth and Atmospheric Sciences at the University of Melbourne, said the work provides scientific evidence for the value of cultural burning and the need to integrate Indigenous knowledge into contemporary fire management. When Country is cared for, he noted, climate does not dictate wildfire. For thousands of years, Gunaikurnai people cared for Country in what is now Croajingolong National Park through cultural burning, keeping much of the landscape as open woodland. The study’s conclusion is that most discussion of Australia’s wildfire crisis stops at climate change, but the new evidence shows what actually set the stage for it: forest expansion and thickening that loaded the landscape with fuel long before the climate began to shift.

The findings carry significant weight for Gunaikurnai Traditional Owners, whose fire practice, known as towera, is understood within their culture as an interrelationship embedded in the practical daily life of the Old Peoples on Country. Uncle Russell Mullett, a Gunaikurnai elder from the Gunaikurnai Land and Waters Corporation, explained that Country not treated with towera becomes extremely dangerous for Gunaikurnai people, animals and plants alike. He welcomed the study as firm scientific evidence supporting towera land management initiatives, giving Traditional Owners a powerful empirical basis for reinstating practices that colonial displacement interrupted. The collaboration between the university researchers and the Gunaikurnai Land and Waters Corporation itself models the kind of partnership the paper argues is essential, combining Western palaeoecological methods with Indigenous ecological knowledge accumulated over millennia.

Beyond its immediate implications for fire policy, the study mounts a direct challenge to one of environmentalism’s most influential ideas: the concept of wilderness. Croajingolong National Park has long been treated as pristine wilderness, land largely untouched by human hands, and that framing has shaped how it is protected. Yet the sediment record shows the opposite. The forest now cherished as natural is, in Professor Fletcher’s words, a product of colonisation, and it is that legacy that now imperils the forest’s very survival. The park is today one of the most fire-prone and repeatedly burnt parts of southeastern Australia, a status the research attributes not to its natural condition but to the absence of the management that once sustained it. Protecting landscapes by excluding the people who shaped them, the study suggests, may be an ecological strategy that actively undermines the ecosystems it aims to preserve.

The technical strength of the work lies in its multi-proxy approach. By combining charcoal morphology to identify fire occurrence, infrared spectroscopy of charred material to reconstruct fire intensity, pollen analysis to track vegetation composition and sediment accumulation rates to quantify erosion, the researchers built a mutually reinforcing picture of cause and effect across a full millennium. The sequencing matters: because the rise in fire temperature precedes the eucalypt expansion in the record, the data support a causal chain in which hotter fires promoted denser forest, denser forest increased fuel loads, and increased fuel loads amplified subsequent fire behaviour. This feedback loop, once initiated by the cessation of cultural burning, converted a fire-resilient open woodland into a fire-vulnerable forest, and it did so before atmospheric greenhouse gas concentrations began their modern rise.

The paper concludes that restoring cultural burning and supporting Traditional Owners to care for Country is an essential component of mitigating wildfire risk, not a cultural gesture but a practical, evidence-based intervention. As southeastern Australian forests face intensifying fire weather under a warming climate, the research suggests that reducing the vulnerability of these landscapes requires addressing the colonial land-use legacies that made them combustible in the first place. The message from Croajingolong is stark and, for fire management agencies, potentially transformative: the current crisis is not solely a climate story, and the tools for rewriting it have been present on Country for thousands of years, waiting to be returned to use.

Subject of Research: Long-term palaeoecological reconstruction linking colonial cessation of Indigenous cultural burning to increased wildfire risk in southeastern Australian forests

Article Title: Australia's wildfire crisis linked to colonial land-use change

Article References: Australia's wildfire crisis linked to colonial land-use change. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: bushfires, cultural burning, Indigenous land management, Gunaikurnai, Croajingolong National Park, eucalypt forests, palaeoecology, lake sediments, colonisation, climate change, fire management, Nature

The central challenge the researchers set out to address is deceptively simple: when a patient is diagnosed with gastric cancer that appears advanced on scans, how reliable is that initial assessment, and what does it actually predict for survival? In Japan, where many tumors that look advanced on imaging turn out to be early-stage disease obscured by inflammatory changes, clinicians have long struggled to identify which patients genuinely benefit from neoadjuvant chemotherapy, or chemotherapy given before surgery. Traditionally, decisions about postoperative chemotherapy have rested on pathological staging, the detailed examination of the removed tumor and lymph nodes. But that approach requires surgery first, leaving no opportunity to shrink the tumor beforehand. With evidence accumulating that perioperative chemotherapy improves survival in resectable advanced disease, the need for trustworthy clinical staging, performed entirely before the operating room, has become urgent.

JCOG1302A was designed as a prospective, multicenter observational study to test exactly how well endoscopy and computed tomography could classify tumor depth and nodal spread. Between July 2013 and November 2014, 1,275 patients were enrolled across 53 specialized Japanese institutions. Each underwent staging with multidetector-row CT using prespecified imaging criteria: localized wall thickening with a smooth serosal surface indicated T2 disease, diffuse thickening with a smooth surface indicated T3, and irregular or nodular serosal change or invasion of surrounding fat signaled T4a, while loss of the fat plane between tumor and adjacent organs defined T4b. Lymph nodes with a short axis of at least 8 millimeters or a long axis of at least 10 millimeters were called metastatic. The earlier analysis found that a striking share of patients clinically labeled T2 or T3 without nodal involvement actually harbored pathologically Stage I disease, contamination that ranged from 50.4 percent in cT2N0 down to just 3.0 percent in cT4aN(+).

That earlier work led the investigators to conclude that patients with clinical T3 disease plus detectable nodes, and those with any T4 disease, were the optimal candidates for preoperative chemotherapy, a conclusion that directly shaped the ongoing Phase III trial JCOG1509. But a critical piece was missing: no study had ever described long-term survival stratified by prospectively determined clinical stage. The new follow-up analysis, JCOG1302A2, collected additional survival data on the original cohort. After excluding patients who did not undergo surgery, those found ineligible, and 83 patients lost to follow-up, 1,176 patients formed the analytical population. The group was overwhelmingly treated with the Japanese standard of gastrectomy plus D2 lymph node dissection, performed in 98 percent of patients, and roughly half received adjuvant chemotherapy, most commonly the oral agent S-1 alone.

The survival data revealed a clean, stepwise gradient across both clinical dimensions. Five-year overall survival was 82.1 percent for clinical T2 tumors, 72.7 percent for T3, 60.0 percent for T4a, and 40.0 percent for the small T4b group. Nodal status told a parallel story: 78.0 percent for node-negative disease, 70.6 percent for cN1, 59.1 percent for cN2, and 28.4 percent for cN3. In statistical models measuring the risk of death over time, clinical T4a carried nearly two and a half times the mortality hazard of T2, while cN3 disease carried almost five times the hazard of cN0. When both dimensions were combined, the picture sharpened further. The most favorable group, comprising cT2N0, cT2N(+), and cT3N0, all achieved five-year survival near or above 77 percent, with cT2N0 topping the chart at 83.5 percent.

Perhaps the most consequential finding concerns the middle of the distribution. Patients with cT3N(+) disease and those with cT4aN0 disease showed nearly identical, intermediate survival, around 68.0 and 66.8 percent at five years respectively, and both carried roughly double the mortality risk of the cT2N0 reference group. Meanwhile, cT4aN(+) patients fared worst among the sizable subgroups, at 57.7 percent. Notably, cT3N0 disease did not show a statistically significant increase in risk compared with cT2N0, suggesting that node-negative T3 tumors, at least when staged by rigorous CT criteria, behave more like earlier disease than like truly advanced cancer. This three-tier risk architecture, favorable, intermediate, and poor, gives clinicians a practical framework for matching treatment intensity to genuine baseline risk.

The results also carry a pointed message about the TNM staging system itself. In the eighth edition of the TNM classification, clinical T3 and T4a tumors are grouped together, despite the acknowledged difficulty of distinguishing them on imaging. The Japanese data, with clearly separated survival curves between the two categories, suggest this merging may obscure clinically meaningful differences. The finding echoes the exploratory radiological sub-analysis of the PRODIGY trial, in which clinical T4 disease, with minimal contamination by pathological Stage I, was identified as the subgroup deriving the greatest benefit from neoadjuvant chemotherapy. Together, these results argue that T3 and T4a disease, when carefully distinguished by multidetector CT, warrant separate consideration in treatment planning and in the design of future clinical trials.

The study’s authors are careful to frame the work as prognostic rather than therapeutic. Because the analysis was observational, it did not directly test whether neoadjuvant chemotherapy improves survival in any subgroup, and the choice and regimen of postoperative chemotherapy were left to treating physicians, varying across patients and institutions. Most patients with Stage III disease received S-1 monotherapy, since the JACCRO GC-07 trial demonstrating the superiority of adding docetaxel had not yet reported when the cohort was treated. Adjuvant chemotherapy use also differed across clinical subgroups, from 34.5 percent in cT2N0 to 65.1 percent in cT4aN(+), a variation that may have influenced the observed outcomes. The cohort excluded scirrhous tumors, bulky nodal disease, and paraaortic metastases, so the findings cannot be generalized to those populations, and all patients were treated at specialized Japanese centers.

Even with those caveats, the timing of this evidence is significant. The Pan-Asian adapted ESMO Clinical Practice Guidelines now emphasize that perioperative chemotherapy should be considered only for selected patients with resectable locally advanced gastric cancer, and the recent MATTERHORN trial has shown that perioperative chemo-immunotherapy can improve outcomes, albeit with greater toxicity and cost. Against that backdrop, knowing which clinically defined groups face the highest baseline mortality risk, and which do well with surgery and standard adjuvant therapy alone, becomes essential for balancing benefit against burden. The data suggest that patients with cT2N0 and cT3N0 disease, who achieved roughly 80 percent five-year survival with upfront surgery, may have little to gain from intensified perioperative regimens, while those with cT3N(+), cT4aN0, cT4aN(+), and cT4b disease represent the higher-risk subgroups in whom such strategies deserve priority investigation.

For a disease that remains one of the world’s leading causes of cancer death, the study delivers something the field has lacked: a prospectively validated, clinically applicable survival map drawn entirely from information available before treatment. It validates the diagnostic criteria used in JCOG1302A, confirms that clinical T and N categories carry genuine prognostic weight, and provides the reference outcomes against which the ongoing JCOG1509 trial and the coming wave of perioperative chemo-immunotherapy studies will be judged. As treatment intensifies, the humble CT scan, applied with disciplined, standardized criteria, may prove to be the most important tool for ensuring that the right patients receive the right therapy at the right time.

Subject of Research: Long-term survival stratified by prospectively determined clinical staging in locally advanced gastric cancer

Article Title: Survival Results by the Prospectively Determined Clinical Staging for Locally Advanced Gastric Cancer: A Follow‐Up Study of JCOG1302A

Article References: Hayashi, T., Kato, R., Mitome, N., Ogawa, R., Fukagawa, T., Katai, H., Nunobe, S., Tokunaga, M., Bando, E., Ito, Y., Yamada, T., Nomura, T., Makino, S., Kinoshita, T., Hara, H., Aizawa, M., Boku, N., Kurokawa, Y., Terashima, M., & Yoshikawa, T. (2026). Survival Results by the Prospectively Determined Clinical Staging for Locally Advanced Gastric Cancer: A Follow‐Up Study of JCOG1302A. Annals of Gastroenterological Surgery, Article ags3.70293. https://doi.org/10.1002/ags3.70293

Image Credits: AI Generated

DOI: 10.1002/ags3.70293

Keywords: gastric cancer, clinical staging, neoadjuvant chemotherapy, JCOG1302A, overall survival, CT imaging, D2 lymphadenectomy, TNM classification, perioperative treatment, oncology, prognosis, Japan Clinical Oncology Group

Rare variant association testing is one of the most powerful strategies in human genetics for connecting genes to disease and to measurable traits such as blood pressure, lipid levels, and kidney function. Because individual rare variants are too uncommon to test one at a time with statistical confidence, researchers aggregate the variants within a gene and ask whether the burden of rare, presumably damaging variants differs between people with high and low trait values. The catch is that the word “presumably” hides an enormous amount of machinery. With tens of thousands of variants per gene and no experimental evidence for most of them, analysts rely on computational annotation methods to decide which variants to count, which to weight heavily, and which to ignore.

That is where the new study intervenes. The team evaluated annotations from five methods spanning three generations of technology: CADD versions 1.6 and 1.7, a long-standing supervised model trained on evolutionary conservation and genomic features; AlphaMissense, a deep learning system from Google DeepMind that predicts the pathogenicity of every possible missense change in the human proteome; ESM-1b, a protein language model that scores variants based on learned patterns of amino acid sequences; and GPN-MSA, a genomic pretrained model that incorporates multiple sequence alignments across species. These five scorers were plugged into four primary gene-based statistical tests and six annotation-level aggregation tests, creating a large grid of method combinations evaluated against the same trait data.

To compare the combinations fairly, the researchers developed a novel evaluation framework based on optimal transport, a mathematical technique originally developed to compare probability distributions and solve problems in logistics and image processing. Applied here, optimal transport allowed the team to quantify how well each annotation method partitioned genetic signal across labeled sets of variants, producing relative measurements of two distinct properties: calibration, meaning whether the test’s statistical behavior matches its theoretical expectations, and power, meaning its ability to detect true associations. Separating these two qualities matters because a test can appear successful simply by being miscalibrated in a way that generates more hits, a trap that has complicated earlier benchmarking efforts in the field.

The results were, by the authors’ own framing, markedly divergent. Tests built on CADD labels achieved the highest signal separation, meaning they did the best job of concentrating genuine genetic signal into the variants flagged as damaging. At the opposite end of the spectrum, tests using AlphaMissense labels showed the lowest calibration, indicating that the statistical behavior of gene-based tests fed with AlphaMissense scores departed most from theoretical expectations, even though AlphaMissense is widely regarded as one of the most accurate single-variant pathogenicity classifiers in clinical interpretation contexts. The lesson is uncomfortable but important: a scorer that excels at classifying individual variants as pathogenic or benign does not necessarily excel at the aggregate, gene-level task of boosting association power.

Perhaps the most visually striking result concerned evolutionary constraint. When the researchers examined which genes were flagged as significant by each annotation method, they found that hits from tests using GPN-MSA labels were strongly enriched among genes with high evolutionary constraint, reaching up to a 5.8-fold enrichment, compared with only 2.1- to 2.8-fold enrichment using the other methods. Evolutionary constraint measures how intolerant a gene is to mutation over millions of years of mammalian evolution, and constrained genes are known to be enriched for genuine disease biology. The GPN-MSA result suggests that a scorer grounded in cross-species sequence alignments directs statistical power toward exactly the genes where rare variant signals are most likely to be real.

Crucially, these differences in variant prioritization measurably shifted the patterns of genetic discovery. Different annotation methods did not merely reshuffle the rankings of the same genes; they changed which genes crossed the threshold of significance at all, meaning that a laboratory’s choice of scorer can determine whether a biologically important gene is reported or missed entirely. Yet the study also found reassuring signs of statistical health across the board: the tests showed low genomic inflation, the standard diagnostic for spurious associations driven by confounding, and similar replication rates when findings were tested in independent data. In other words, all the method combinations were behaving statistically responsibly; they were simply looking for different things.

The authors distill their findings into three conclusions with direct practical consequences for the field. First, no single combination of annotation method and statistical test is currently optimal for gene-level association testing, so there is no defensible default that a study can adopt without justification. Second, results from different methods and tests can and should be aggregated to increase study power, since the methods appear to capture partially overlapping but non-identical slices of the underlying biology, and combining them recovers signal that any one approach loses. Third, integrating the information content of variant annotations and gene-level annotations remains a key open problem, a gap that the authors identify as the central methodological challenge facing rare variant genetics.

The scale of the underlying data lends the conclusions considerable weight. The analysis drew on the UK Biobank under application number 44257 and additionally used data from the National Institutes of Health’s All of Us Research Program, with the authors acknowledging the hundreds of thousands of participants whose genomes and health measurements made the comparison possible. All authors were employees of Genentech at the time the work was performed, and several hold Roche stock, a disclosure that situates the study within an industrial research environment where rare variant discovery pipelines feed directly into drug target identification. For pharmaceutical companies, a method that systematically misses constrained genes is not an abstract statistical concern; it is a missed therapeutic hypothesis.

For the broader community, the study arrives at a moment when machine learning scorers are proliferating faster than the field can benchmark them. AlphaMissense, ESM-1b, and GPN-MSA represent a new wave of models built on deep learning architectures originally developed for natural language, and each embodies different assumptions about what makes a variant harmful: supervised training on known disease mutations, learned protein sequence grammar, or evolutionary conservation across species. The new results demonstrate that these assumptions translate into genuinely different discovery profiles at scale, not merely different numbers on a leaderboard. Researchers planning gene-based association studies should treat annotation choice as a first-order experimental design decision, ideally running multiple scorers and aggregating results, and method developers should recognize that the gene-level aggregation setting, not just single-variant classification accuracy, is where their tools will ultimately be judged. The full study is openly available under a Creative Commons license, with the DOI 10.1186/s12864-026-13379-2.

Subject of Research: Comparative performance of machine learning variant annotation methods in gene-level rare variant association testing

Article Title: Markedly divergent performance of variant annotation methods for gene-level association testing

Article References: Aguirre, M., Irudayanathan, F. J., Crow, M., Hejase, H. A., Menon, V. K., Pendergrass, R. K., McCarthy, M. I., & Fletez-Brant, K. (2026). Markedly divergent performance of variant annotation methods for gene-level association testing. BMC Genomics. https://doi.org/10.1186/s12864-026-13379-2

Image Credits: AI Generated

DOI: 10.1186/s12864-026-13379-2

Keywords: variant annotation, gene-level association testing, rare variants, UK Biobank, CADD, AlphaMissense, ESM-1b, GPN-MSA, optimal transport, genetic association study, machine learning, genomics