Find the papers that actually matter

Molecular profiling of gliomas has a pivotal role in diagnosis, treatment selection, and prognostic assessment. However, it heavily relies on time-consuming and expensive genomic testing, which is largely inaccessible in resource-limited settings. To enable cost-effective and scalable identification of molecular alterations, we developed and validated a foundation model-based interpretable approach to predict key molecular events directly from routine histopathology slides without manual annotation. We developed the glioma molecular alterations predictor (GMAP), a foundation model-based approach using 1696 whole-slide images from 877 patients downloaded from the Cancer Genome Atlas. The model was validated on an internal test set (167 whole-slide images from 88 patients) and a grouped external validation set (4602 whole-slide images from 3147 patients; 12 Chinese hospitals and a public dataset, EBRAINS). The performance was primarily evaluated at the patient level by the area under the receiver operating curve (AUROC), accuracy, sensitivity, specificity, and F1 score, with probabilities aggregated across multiple slides per patient by averaging. The interpretability was evaluated through multilevel analysis of high-contribution tiles, and comparative assessment between model-generated heatmaps and corresponding immunohistochemical staining patterns. The GMAP reached AUROCs of 0·939 (95% CI 0·865-0·993) for isocitrate dehydrogenase (IDH), 0·955 (0·898-0·992) for the co-deletion of chromosome arms 1p and 19q (1p/19q co-deletion), 0·944 (0·849-1·000) for telomerase reverse transcriptase (TERT), and 0·886 (0·802-0·955) for chromosome 7 gain and chromosome 10 loss (+7/-10) on the internal test set, respectively. In the grouped external validation set, the AUROCs was 0·870 (95% CI 0·857-0·883) for IDH, 0·885 (0·865-0·905) for 1p/19q co-deletion, 0·694 (0·665-0·724) for TERT, and 0·672 (0·615-0·727) for +7/-10. Interpretability analysis showed that GMAP attends to both known and previously unrecognised morphological characteristics associated with molecular alterations. GMAP offered a technically feasible approach for accurate, fast, and potentially cost-effective identification of molecular alterations in resource-constrained settings. Interpretability analysis revealed model-attended features, which improve the model's trustworthiness for clinical adoption. National Natural Science Foundation of China.

Prognostic stratification in stage III colon cancer remains poor, despite treatment advances. Tumor-infiltrating lymphocytes, particularly CD3+ T cells, are potential prognostic markers, but manual assessment is labor-intensive and not robust. This study aimed to develop a deep learning model for automated analysis of CD3-stained histological slides to improve prognostic prediction. A total of 1737 patients from three international cohorts (PETACC08, PRODIGE-13, and HARMONY) were analyzed. The deep learning model (VGG19) identified tumor core (TC) and invasive margin (IM) regions on CD3-stained slides. Features from VGG19 and UNI models were used to cluster patients using hierarchical classification. Prognostic performance was evaluated using disease-free survival (DFS) across training, internal validation, and external validation sets. Deep learning classifiers identified distinct patient clusters with significantly different DFS based on TC and IM. For both IM and TC analysis, patients in the favorable group had a better DFS in all sets (IM: p < 0.001, p = 0.04, p = 0.02; TC: p = 0.002, p = 0.01, p = 0.12, respectively). Combining classifiers enhanced prognostic accuracy in all sets (p < 0.001, p = 0.01, p = 0.06, respectively). The model outperformed traditional clinical variables and CD3 enumeration, which demonstrated variability across cohorts. Automated deep learning analysis of CD3-stained slides enables robust and reproducible prognostic stratification in stage III colon cancer, independently of staining and scanning variations. This approach holds promise for guiding personalized treatment strategies.ClinicalTrials.gov Identifiers: NCT00265811, NCT00995202.

Immune checkpoint inhibitors (ICIs) have transformed the treatment of advanced melanoma, yet their efficacy is limited by high-grade immune-related adverse events that often require treatment with systemic corticosteroids. Although corticosteroids are widely used, their impact on anti-tumor immunity remains poorly defined. Using an ICI-responsive murine melanoma model, we show that tapered systemic prednisolone administered after three cycles of combined anti-CTLA4 and anti-PD1 therapy compromises ICI-mediated tumor control, leading to delayed progression in one-third of initially responding animals. Mechanistically, prednisolone selectively suppressed CD8+ effector T-cell activation in tumor-draining lymph nodes and in the circulation, while expanding activated regulatory T-cells. These changes increased the Treg:CD8+ effector ratio, reduced cytotoxic T-cell function and blocked the early ICI-mediated induction of cytokines, including IL-2, IFNγ, VEGF, CCL3/4, IL-13, IL-3, and GM-CSF. Importantly, despite these early immunosuppressive effects, long-term tumor-specific memory responses were preserved. Autologous melanoma:T-cell cocultures validated these findings. Overall, systemic prednisolone disrupts early CD8+ T-cell-mediated anti-tumor activity but spares durable immunity, highlighting the critical importance of timing and context in the introduction of corticosteroids during ICI therapy.

Prostate cancer (PCa) is a leading cause of male cancer-related death globally. While the gut microbiota is linked to PCa, its genetic association remains unclear. We screened genetic instruments related to the gut microbiota and paired them with PCa genome-wide association study data to conduct Mendelian randomization (MR) analysis. Positive MR findings were then subjected to colocalization analysis. Subsequently, we utilized the Gene Expression Omnibus (GEO) dataset to perform differential expression analysis, aiming to identify differentially expressed associated genes (DEAGs). We determined the importance scores of these DEAGs through four machine learning models and constructed a nomogram based on these findings, and then validated it in another group of the GEO dataset. MR analysis found 16 gut bacteria causally linked to PCa (7 risk, 9 protective), with 144 related genes. PLCL1, VSNL1, ROR2, NRXN3, and TEAD1 were identified as feature genes for constructing a nomogram that provides a quantitative prediction of the risk of PCa onset. This study indicates that there are causal links between the gut microbiota and PCa. Feature genes may affect the occurrence of PCa by inhibiting the epithelial-mesenchymal transition, proliferation, migration, and invasion of cells.

Iron chelation therapy titration is managed by hematologists who monitor iron levels and adjust medication dosages to achieve patient outcomes. This study developed a machine learning (ML) algorithm to predict the liver iron concentration (LIC) and chelation therapy adjustments. This retrospective, single-centre cohort study included adult patients who underwent FerriScan MRI to obtain LIC for chelation therapy monitoring. Using an XgBoost-based ML framework, the proximal-time model (PTM) utilised clinical/drug, laboratory and MRI data features from one visit prior to the target visit, whereas the all-time model (ATM) utilised the data from all prior visits. 94 patients with 892 consecutive visits between January 2008 and November 2018 were included in this study. We assessed the prediction capabilities of the PTM and ATM in LIC, changes to chelation drug type and dosage changes. The PTM model was superior to the ATM model in all the experiments. When drug features were excluded, the CLICT model for predicting patient iron overload status improved to an AUROC of 0.83 [95% CI 0.75-0.91] for PTM; compared to an AUROC of 0.73 [95% CI 0.66-0.80] when drug features were included.For predicting changes in chelation type, the CLICT model showed AUROC of 0.83 [95% CIs 0.77-0.89] for PTM. There is high concordance of the agreement of hematologists with ML in not changing the chelation drug type. The ML model is a step toward creating a clinical decision support system tool for the prediction of LIC and iron chelation therapy adjustment in patients with haemoglobinopathies or hemolytic anemias.

Esophageal adenocarcinoma (EAC) has seen a dramatic rise in incidence in developed countries over the past three decades. Early detection of its precursors-gastroesophageal reflux disease (GERD), Barrett's esophagus (BE), and high-grade dysplasia (HGD) is critical for cancer prevention. This study presents the development and validation of a novel liquid biopsy assay based on circulating microbial DNA (cmDNA) for the early detection of EAC and HGD. Using metagenomic sequencing, we identified significant differences in microbial diversity and composition between EAC and HGD patients, as well as between BE and GERD patients. A total of 46 microbial candidates in tissue and 419 in serum were upregulated in EAC & HGD, with 11 consistently elevated in both sample types. Following qRT-PCR validation and LASSO regression, a 6-marker cmDNA panel was selected. This signature was incorporated into a diagnostic model trained with the XGBoost algorithm, achieving an AUC of 0.93 in the training cohort (52 HGD & EAC cases vs. 54 BE & GERD controls). Importantly, the model demonstrated robust performance in an independent testing cohort (23 HGD & EAC cases vs. 22 BE & GERD controls), yielding AUCs of 0.91 for EAC and 0.88 for HGD. These findings highlight the diagnostic potential of cmDNA-based profiling and support its utility as a minimally invasive, accurate, and generalizable tool for early detection of esophageal adenocarcinoma.

To construct and interpret a machine learning model for predicting overall survival in nonsurgical prostate cancer with bone metastases (PCBM). Data from 3,378 SEER database patients were utilized to develop machine learning survival models, with the best-performing model visually interpreted using SHAP. The Extra Survival Trees (EST) model performed best (validation AUC = 0.694, C-index = 0.643). SHAP analysis identified the Gleason score as the most critical survival factor, significantly outweighing clinical T stage. Visceral metastasis and advanced age also markedly increased mortality risk. The EST model effectively assesses OS in nonsurgical PCBM. The Gleason score holds greater prognostic value than local anatomical staging in this cohort, suggesting clinicians should prioritize early, aggressive combination treatments for high-Gleason, high-burden patients.

Increasing evidence suggests that Graves' disease (GD) may increase the risk of thyroid cancer (THCA), but diagnostic biomarkers associated with it remain underexplored. To address this issue, we analyzed the Gene Expression Omnibus (GEO) and TCGA (The Cancer Genome Atlas) databases, identified 21 shared immune-related genes via differential expression analysis and weighted gene coexpression network analysis (WGCNA). Gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses emphasized immune pathways, and LASSO regression was used to select five core genes (TREM1/S100A11/MRPS16/MET/ACTN1) to construct robust diagnostic models. The CIBERSORT algorithm revealed a significant correlation between the models and immune infiltration of the THCA. Machine learning and protein‒protein interaction (PPI) networks revealed TREM1 as a central gene for predicting the response to immunotherapy. Xenograft tumor models confirmed that TREM1 knockdown suppressed the proliferative capacity of thyroid cancer cells in vivo. Drug sensitivity studies identified VER-155008 as a potential therapeutic compound. Bioinformatics and experimental validation (qRT‒PCR) revealed that the HOTTIP/hsa-miR-204-3p/TREM1 axis serves as a ceRNA to regulate TREM1. Our study identified five core genes, with TREM1 as a central regulator, that demonstrate strong diagnostic potential for both Graves' disease (GD) and thyroid carcinoma (THCA). These findings provide valuable diagnostic biomarkers and therapeutic targets for THCA patients with GD.

Targeting programmed cell death protein 1 (PD-1) and cytotoxic T-lymphocyte-associated protein 4 (CTLA-4) with immune checkpoint inhibitors (ICIs) has improved survival across multiple cancer types, but the variability in patient response highlights the need for better predictive biomarkers. Existing studies rely on taxonomic abundance derived from reference genome databases, limiting the discovery and functional interpretation of uncharacterized microbes. Here, we integrated metagenomic data from multiple ICI-treated cohorts spanning diverse cancer types and geographic regions and developed a deep learning model, named BioP-VAE, that incorporates biological prior knowledge via protein sequence embeddings and uses gene-level microbial abundance features as input. Gene-level microbial abundance outperformed taxonomy abundance in predicting both ICI response and 12-month progression-free survival (PFS). In patients receiving combination immune checkpoint blockade (CICB), BioP-VAE achieved a mean AUC of 0.89 in intracohort and 0.88 in cross-cohort evaluation. Notably, in the monotherapy-treated intracohorts, BioP-VAE achieved a mean AUC of 0.97. Feature attribution analysis revealed key microbial genes. Additionally, we identified distinct predictive microbial signatures via age-stratified analysis, suggesting that host age may modulate microbiome‒immune interactions. Importantly, this is the first large-scale study to evaluate gene-level microbial abundance features for ICI response prediction across multiple cancer types by deep learning. Our findings demonstrate that incorporating biological prior knowledge into deep learning models can improve the discovery of microbial biomarkers that can be generalized across cancer types and treatment settings, offering a novel strategy for patient stratification in immunotherapy.

The selection of an optimal maintenance agent in diffuse large B-cell lymphoma (DLBCL) continues to pose a significant clinical challenge. This study aims to evaluate the prognostic impact of maintenance therapy (MT) in DLBCL. We conducted a retrospective analysis of data from DLBCL patients undergoing first-line MT at four hospitals in Beijing between January 2019 and August 2024. The REMoDL-B trial database was selected as the control group. The MT group comprised 106 cases and a median follow-up duration of 25.4 months. The rates of progression and death were 11.32% (12/106) and 1.89% (2/106), respectively. The 2-y progression-free survival (PFS) and overall survival (OS) rates were 90% and 98%, respectively. The MT group demonstrated significantly superior PFS and OS compared to the control group (p = 0.024, p = 0.008). Furthermore, multivariate analysis indicated that MT (p = 0.021, OR = 0.037, 95% CI, 0.002-0.605) was an independent prognostic factor associated with improved PFS. For patients receiving Bruton tyrosine kinase inhibitors (BTKi), the 2-y PFS and OS rates were 87.6% and 97.2%, respectively, both significantly better than those of the control group (p = 0.048, p = 0.024). Despite 43.6% of patients being at high risk for central nervous system (CNS), no CNS recurrences were observed. The PFS of the MCD subtype is better than that of the A53 subtype. While limited by the retrospective study, our analysis raises the hypothesis that MT may correlate with improved DLBCL outcomes. A similar trend suggesting potential benefit from BTKi maintenance was noted, meriting further investigation in controlled settings.