AI and Cancer Research ── 2026-11-17
New papers on cancer and AI, collected from journal feeds, PubMed (searched across all journals) and OpenAlex (a public index used instead of Google Scholar), judged by Jev (an AI that makes language judgements): original research, AI central, which of 13 themes, treatment or trials, and then ranked by importance up to a daily limit. Journals are not filtered by impact factor; each paper shows its journal's IF-equivalent (OpenAlex two-year mean citedness) and the date of that value. Each entry shows the full original abstract.
All cancers: Cancer biology(2)
Hematologic malignancies: AML(0) · CML(0) · ALL(0) · CLL(0) · Lymphoma(0) · Myeloma(0)
Solid tumors: Breast(2) · Lung(1) · Colorectal(1) · Pancreatic(0) · Prostate(0) · Brain(0) · Ovarian(0) · Endometrial(0) · Gastric(1) · Liver(1) · Kidney(1) · Bladder(1)
Rare cancers: ACC(0)
Cancer molecular biology and AI
- Abstract
Heterogeneity in tumor regulatory programs, cellular composition and spatial organization complicates target prioritization for natural compounds. We developed a spatial-context heterogeneous graph neural network ranker (SC-HGNN-Rank) that combines relation-aware graph representations with ranking optimization to integrate compound-target associations, transcriptional regulation, interaction networks, and single-cell and spatial information. The model ranked 600 candidate axes in colon adenocarcinoma (COAD), hepatocellular carcinoma (LIHC) and gastric adenocarcinoma (STAD). Under matched training conditions, mean Spearman correlations with the reference ranking were 0.965, 0.967 and 0.976 for leave-one-cancer-out, compound-group holdout and target-group holdout evaluations, respectively. The composite ranking objective improved mean Spearman correlation by 0.049 over a simple objective using the same encoder, whereas removing the spatial module reduced it by 0.147. Model ranking and multilayer evidence prioritized Quercetin-CDK1/MYB-Myeloid in COAD, 4',5-dihydroxyflavone (4',5-DHF)-AR/HDAC1-T/NK in LIHC, and Eriodyctiol-HSP90AB1/HOPX-Fibroblast in STAD. Expanded spatial analyses with expression-matched gene controls assessed associations between target expression and mapped cell abundance. Cellular thermal shift and drug affinity responsive target stability assays showed increased retention of target proteins after compound treatment. In HCT116, JHH-6 and MKN45 cells, the compounds reduced viability, colony formation, migration and invasion cell counts, and increased apoptosis. Target knockdown produced similar growth-inhibitory phenotypes, whereas overexpression attenuated compound-induced effects, supporting the involvement of CDK1, AR and HSP90AB1 in the respective responses. The accompanying Explorer supports candidate searches, evidence inspection and data export. By incorporating regulatory and spatial contexts into natural-compound prioritization, this study provides a reusable approach and experimentally supported candidates for target-directed pharmacological research in digestive system cancers.
Journal IF-equivalent: 11.5 (OpenAlex 2-year mean citedness, value as of 2026-11-16, retrieved 2026-11-17; not the official Clarivate IF)Reference: Jiang H, Zheng S, Yang C, Zhang X, Liu H, Bao Y. Graph learning guided by regulatory and spatial contexts prioritizes natural compound–target pairs in digestive system cancers. Pharmacol Res. 2026 Nov [Epub ahead of print]. doi:10.1016/j.phrs.2026.109767. PMID: 43035199.Checked: Abstract only - Abstract
Accurate peptide-HLA class I binding prediction is fundamental for identifying tumor neoantigens and advancing personalized immunotherapy. We present LKMHCpan, a pan-specific predictor that leverages transformer networks to optimize HLA-I pseudo-sequences representation through attention-guided integration of 88 key residues primarily within the α1, α2, and α3 domains. LKMHCpan achieves a mean precision-recall area under the curve (PR-AUC) of 0.93, generalizes robustly to untrained HLA-I alleles, and supports peptides longer than 11 amino acids. In Tumor Neoantigen Selection Alliance (TESLA) data, LKMHCpan identified 94% of immunogenic neoantigens with strong score-immunogenicity correlation. Neoantigen burden from LKMHCpan significantly stratified clinical outcomes in two immunotherapy-treated cohorts: metastatic melanoma ( n = 73) and lung cancer ( n = 309), with high-burden patients exhibiting clinical benefit. Its ability to resolve binding motifs for untrained alleles and single-amino-acid HLA-I mutants highlights its potential for elucidating complex peptide-major histocompatibility complex (MHC) interactions. LKMHCpan provides a scalable framework for neoantigen discovery and prognostic assessment in precision immunotherapy.
Journal IF-equivalent: 5.3 (OpenAlex 2-year mean citedness, value as of 2026-11-16, retrieved 2026-11-17; not the official Clarivate IF)Reference: Li Q, Liu S, Yao M, Chen H, Yu Y, Zhang B, et al. LKMHCpan: Transformer-optimized HLA-I pseudo-sequences enhance neoantigen prediction for immunotherapy. iScience. 2026 Nov;29(11):118134. doi:10.1016/j.isci.2026.118134. PMID: 43034688.Checked: Abstract only
See all 2 Cancer biology papers →
Hematologic malignancies
Acute myeloid leukemia (AML): treatment, trials and AI
No new qualifying paper for this issue.
Chronic myeloid leukemia (CML): treatment, trials and AI
No new qualifying paper for this issue.
Acute lymphoblastic leukemia (ALL): treatment, trials and AI
No new qualifying paper for this issue.
Chronic lymphocytic leukemia (CLL): treatment, trials and AI
No new qualifying paper for this issue.
Lymphoma: treatment, trials and AI
No new qualifying paper for this issue.
Multiple myeloma: treatment, trials and AI
No new qualifying paper for this issue.
Solid tumors
Breast cancer: treatment, trials and AI
- Abstract
PURPOSE: Suboptimal postoperative functional exercise adherence remains prevalent among breast cancer patients, posing a major barrier to effective rehabilitation in nursing practice. This study aimed to develop and internally validate an interpretable machine learning model to identify patients with poor postoperative functional exercise adherence, examine the relative contribution of illness perception to model classification, and support risk-stratified rehabilitation nursing interventions. METHODS: A cross-sectional study enrolled 904 postoperative breast cancer patients at a tertiary hospital in Anhui between January 2024 and January 2026. Participants were randomly divided into training and internal validation sets at a 7:3 ratio to develop and evaluate seven machine learning models, with LASSO-based feature selection and Optuna hyperparameter tuning using 10-fold cross-validation. Functional exercise adherence and illness perception score were assessed using validated scales. Shapley additive explanations (SHAP) interpreted the optimal model. RESULTS: In the training set, 59.4% of patients demonstrated good postoperative functional exercise adherence. Age, body mass index (BMI), radiotherapy, postoperative time, education and illness perception were independent predictors, with higher illness perception scores associated with lower odds of poor adherence (OR = 0.434). Among candidate models, the Light Gradient Boosting Machine (LightGBM) model achieved the best overall performance. The model showed good discrimination and calibration in the training set (AUC = 0.874; sensitivity = 0.819; specificity = 0.723), with comparable performance observed in the validation set (AUC = 0.839; sensitivity = 0.890; specificity = 0.626). SHAP analysis identified illness perception score as the most influential feature in the LightGBM model, highlighting its contribution to interpretable adherence classification. CONCLUSION: An interpretable machine learning model was developed and internally validated to identify poor postoperative functional exercise adherence among breast cancer patients. Illness perception emerged as the most influential feature in the model, highlighting its potential importance in adherence screening and risk-stratified rehabilitation nursing. Postoperative functional exercise adherence remains suboptimal in breast cancer patients. An interpretable prediction model was developed to identify patients at risk of poor adherence. Illness perception emerged as a key modifiable predictor of adherence. The model supports early risk identification and risk-stratified rehabilitation nursing.
Journal IF-equivalent: 3.5 (OpenAlex 2-year mean citedness, value as of 2026-11-16, retrieved 2026-11-17; not the official Clarivate IF)Reference: Yang G, Cui S, Wu J, Wu Z, Liu J, Fan B, et al. Development and Validation of an Interpretable Machine Learning Model for Predicting Postoperative Functional Exercise Adherence in Breast Cancer Patients: The Central Role of Illness Perception. Patient Prefer Adherence. 2026 Nov;Volume 20:1-18. doi:10.2147/ppa.s653378. PMID: 43036587.Checked: Abstract only - Abstract
BACKGROUND: Radiotherapy-induced lymphopenia (RIL) is a clinically relevant hematologic toxicity in patients with breast cancer, yet early risk assessment before radiotherapy remains challenging because existing approaches often rely on dosimetric parameters or hematologic changes observed during treatment. This study aimed to develop and internally validate an interpretable machine-learning (ML) model using routinely available pre-radiotherapy clinical and laboratory variables to estimate the risk of grade ≥2 RIL in patients with breast cancer. METHODS: We retrospectively studied 167 patients with breast cancer treated at Siyang Hospital between 2020 and 2025, including 56 with grade ≥2 RIL. Six predictors preselected from 20 routine pre-radiotherapy variables were used to compare seven algorithms through repeated nested fivefold cross-validation (20 repetitions; fivefold inner tuning). Performance was evaluated using patient-level mean out-of-fold predictions. The final model was interpreted using Shapley additive explanations and implemented as a research calculator. Irradiation extent and dosimetric parameters were excluded. RESULTS: Random forest achieved an area under the receiver operating characteristic curve of 0.801 (95% confidence interval [CI]: 0.722-0.871), an area under the precision-recall curve of 0.750 (95% CI: 0.647-0.842), and a Brier score of 0.159 (95% CI: 0.128-0.193). Sensitivity was 60.7% and specificity 89.2%; this moderate sensitivity limits identification of patients who develop grade ≥2 RIL. The final model included age, absolute neutrophil count, absolute lymphocyte count, absolute monocyte count, lactate dehydrogenase, and prealbumin. CONCLUSION: This internally validated model showed moderate discrimination. Early predictor preselection may inflate performance estimates, and independent external validation is required. The calculator remains a research tool and should not guide clinical decisions.
Journal IF-equivalent: 2.9 (OpenAlex 2-year mean citedness, value as of 2026-11-16, retrieved 2026-11-17; not the official Clarivate IF)Reference: Ding L, Qiu H, Wang M, Lu Y, Huang T, Wang Y, et al. Pre-Radiotherapy Risk Prediction of Grade ≥2 Radiotherapy-Induced Lymphopenia in Breast Cancer: A Single-Center Interpretable Machine Learning Study. Int J Gen Med. 2026 Nov;Volume 19:1-13. doi:10.2147/ijgm.s655604. PMID: 43032615.Checked: Abstract only
Lung cancer: treatment, trials and AI
- Abstract
PURPOSE: Given the institutional and regulatory barriers to centralized sharing of multicenter whole-slide image (WSI) data for non-small cell lung cancer (NSCLC), federated learning (FL) enables collaborative model development without requiring centralized sharing raw WSI data. This study explores a WSI-based FL model for predicting chemoradiotherapy response in advanced NSCLC, and compares its performance with centralized deep learning (DL) and clinical models using external multi-center validation cohorts. METHODS: This study collected WSI and corresponding clinical data from 250 patients across four hospitals. We developed a centralized DL model and a FL model based on WSI, with patch-level predictions aggregated to patient-level outcomes using a voting method. In addition, we built a clinical logistic regression model using clinical data. All models were evaluated on two independent external cohorts (Hospital C and Hospital D). Performance was summarized as the mean and 95% confidence interval across the five models obtained from five-fold cross-validation. RESULTS: Across both external validation cohorts, WSI-based models consistently outperformed the clinical model. The FL model achieved performance comparable to the DL model. In hospital C, the DL model had an accuracy/AUC of 0.68/0.64, while the FL model achieved 0.70/0.65. The clinical model had an accuracy/AUC of 0.58/0.58. In hospital D, the DL model had an accuracy/AUC of 0.68/0.68, while the FL model achieved 0.73/0.71. The clinical model had an accuracy/AUC of 0.61/0.59. CONCLUSION: The FL model using WSI achieved performance comparable to the centralized DL model in predicting NSCLC outcomes, while allowing each participating institution to retain its local data. It also showed consistent performance across independent external validation cohorts. This approach provides a practical solution for collaborative model development in multi-center pathology studies, particularly in settings where data sharing is restricted.
Journal IF-equivalent: 4.8 (OpenAlex 2-year mean citedness, value as of 2026-11-12, retrieved 2026-11-13; not the official Clarivate IF)Reference: Liu Y, Pan Y, Tian Y, Zhang S, Liu Q, Qiu J, et al. A decentralized federated learning approach for predicting chemoradiotherapy response from multicenter whole-slide images. BMC Cancer. 2026 Aug 19;26(1):1422. doi:10.1186/s12885-026-16716-6. PMID: 43035727.Checked: Abstract only
Colorectal cancer: screening, treatment, trials and AI
- Abstract
Newest evidence shows a link between early-onset colorectal cancer (EO-CRC) and metabolic dysfunction-associated steatohepatitis (MASH). The molecular mechanism behind this connection is not clear. This study tries to find key biomarkers linked to the progression and treatment response of MASH-associated EO-CRC. Differential expression analysis (DEGs) and weighted gene co-expression network analysis (WGCNA) were performed to mine associated genes. Forty-four machine learning algorithms were used to screen candidate biomarkers. For the purpose of pinpointing critical biomarkers, this study, integrating survival analysis, expression profiling, and prognostic assessment, was carried out. Through the application of single-cell RNA sequencing (scRNA-seq) data and in vitro experimental approaches, verification, of gene expression and functional roles, was achieved. With the edgeR package and WGCNA, coexpressed genes enriched in matrix remodeling and lipid metabolism pathways are obtained, and seven candidate biomarkers are selected by machine learning, among which only FADS1 connects to disease prognosis and advanced stage. scRNA-seq and virtual knockout experiments revealed that FADS1 in tumor cells is associated with fibrosis and angiogenesis. Further validation using quantitative real-time PCR (qRT-PCR), Western blotting and immunohistochemistry (IHC) staining identified FADS1 as a candidate shared biomarker, offering a promising theoretical basis for fundamental tumor research and clinical translational applications.
Journal IF-equivalent: 3.3 (OpenAlex 2-year mean citedness, value as of 2026-11-16, retrieved 2026-11-17; not the official Clarivate IF)Reference: Xia H, Hou Y, He X, Yu Q, Zhang X, Zhou Y. FADS1 serves as a candidate shared biomarker for metabolic dysfunction‑associated steatohepatitis related early‑onset colorectal cancer identified via machine learning and in vitro validation. Discov Onc. 2026 Nov 15 [Epub ahead of print]. doi:10.1007/s12672-026-06235-y.Checked: Abstract only
Pancreatic cancer: treatment, trials and AI
No new qualifying paper for this issue.
Prostate cancer: treatment, trials and AI
No new qualifying paper for this issue.
Brain tumours: treatment, trials and AI
No new qualifying paper for this issue.
Ovarian cancer: treatment, trials and AI
No new qualifying paper for this issue.
Endometrial cancer: treatment, trials and AI
No new qualifying paper for this issue.
Gastric cancer: treatment, trials and AI
- Abstract
BACKGROUND: Immunotherapy has become a promising strategy for gastric cancer, yet predicting treatment response remains challenging. We developed an exploratory, biologically informed machine learning framework integrating peri-treatment serum biomarkers and clinical variables, with candidate cytokines informed by an in vitro patient-derived organoids (PDOs)-cancer-associated fibroblasts (CAFs)-T cells co-culture platform, to assess pathological response to PD-1 inhibitor (PD-1i)-based therapy in gastric cancer. METHODS: Data from 125 patients receiving PD-1i were analyzed. Serum samples collected before and after immunotherapy were tested for cytokines. Eighteen successfully established PDOs-CAFs-T cells co-culture models were used for cytokine profiling and to identify seven candidate cytokines. These cytokines were subsequently measured in paired serum samples from the 125-patient clinical cohort and combined with clinicopathological variables for machine learning model development. Tumor Regression Grade (TRG) served as the binary outcome variable, representing patients with good versus poor/no tumor regression. Experimental and clinical data were used to train predictive models with multiple algorithms, including CatBoost and several other machine learning methods. Model performance was assessed by tenfold cross-validation, ROC, precision-recall (PR) Curve, and decision curve analyses. Feature importance was interpreted using Shapley additive explanations (SHAP) and visualized using decision tree. RESULTS: The PDOs-CAFs-T cells co-culture platform showed cytokine-change trends that were generally concordant with those observed in patient serum after treatment. 10 key predictors were identified, including pre-/post-treatment maximum diameter, ycT/ycN stages, and pre- or post-treatment cytokines (IFN-α2, IL-2, IL-6, IL-8, MCP-1). In the internal testing set, the tuned CatBoost model achieved an AUC of 0.845. SHAP analysis revealed post-treatment maximum diameter and cytokine profiles as the most influential features. Decision trees clearly and interpretably present the model. CONCLUSION: This study presents an exploratory, biologically informed machine learning framework for peri-treatment assessment of pathological response in gastric cancer. The PDOs-CAFs-T cells co-culture platform provided biological context for candidate cytokine selection.
Journal IF-equivalent: 6.2 (OpenAlex 2-year mean citedness, value as of 2026-11-16, retrieved 2026-11-17; not the official Clarivate IF)Reference: Chen R, Geng T, Xiao P, Zhang Y, Wang X, Yu Z, et al. Interpretable machine learning for predicting immunotherapy response in gastric cancer via serum cytokine profiling and organoid co-culture platforms. Cancer Immunol Immunother. 2026 Aug 19;75(11):277. doi:10.1007/s00262-026-04509-6. PMID: 43035594.Checked: Abstract only
Liver cancer: treatment, trials and AI
- Abstract
OBJECTIVE: To develop and internally validate an interpretable machine learning model for predicting acute chemotherapy-induced nausea and vomiting (ACINV) after intra-arterial interventional therapy (IAIT) in patients with primary liver cancer. METHODS: This retrospective cohort included 1062 patients who underwent IAIT in China. ACINV was defined as Common Terminology Criteria for Adverse Events (CTCAE) grade ≥ 2 nausea or vomiting within 24 hours. Patients were allocated to a training set ( n = 638) and a held-out internal validation set ( n = 424). Seven algorithms were compared using nested stratified five-fold cross-validation. Outer folds generated out-of-fold (OOF) predictions, whereas inner folds supported tuning and threshold selection. Imputation, encoding, standardization, and least absolute shrinkage and selection operator (LASSO) selection were fitted only to training observations in each partition. Preprocessing was re-estimated within folds used for LASSO penalty selection. SMOTE was applied only within model-fitting partitions after predictor selection. Algorithms were ranked by pooled OOF area under the curve (AUC), with the Brier score resolving ties. The selected pipeline was refitted on the complete training set and evaluated in internal validation. RESULTS: ACINV occurred in 380 patients (35.8%). The final refit retained 11 predictors. Random forest achieved the highest pooled OOF AUC (0.941; 95% CI, 0.923-0.958) and a Brier score of 0.092. In internal validation, its AUC was 0.962 (95% CI, 0.946-0.978), Brier score was 0.073, calibration intercept was -0.090, and calibration slope was 1.056. SHAP analysis supported model interpretation, and the locked model was incorporated into a prototype calculator. CONCLUSIONS: The model showed promising internal performance. External validation and prospective impact evaluation are required before clinical implementation. TRIAL REGISTRATION: This study was registered in Chinese Clinical Trial Registry (ChiCTR-2500110585).
Journal IF-equivalent: 2.8 (OpenAlex 2-year mean citedness, value as of 2026-11-16, retrieved 2026-11-17; not the official Clarivate IF)Reference: Lu Y, Zheng R, Yu H, Yang X, Liu Y, Wang Y, et al. Development and internal validation of a machine learning model for predicting acute chemotherapy-induced nausea and vomiting after intra-arterial interventional therapy in patients with liver cancer. Asia Pac J Oncol Nurs. 2026 Dec;13:101040. doi:10.1016/j.apjon.2026.101040. PMID: 43034587.Checked: Abstract only
Kidney cancer: treatment, trials and AI
- Abstract
BACKGROUND: Treatment selection for localized renal cell carcinoma (RCC) has progressed significantly in recent years. OBJECTIVES: We aim to describe clinicopathological characteristics and treatment outcomes of clinical stage I RCC patients, and construct machine learning (ML) models for personalized prediction of survival outcomes at various epochs pre- and post-operatively. DESIGN: This is a single-center retrospective study of patients treated for clinical stage I RCC from July 1990 to December 2023. Exclusion criteria included patients with stage II-IV RCC at diagnosis, patients who were not treated (e.g. active surveillance, basic supportive care), and incomplete data. METHODS: ML models using extreme gradient boosting were created to predict overall survival (OS), recurrence-free survival (RFS), cancer-specific survival (CSS), and postoperative complications in clinical stage I RCC, with evaluation via concordance-index (C-index) for survival and accuracy metrics for complications. RESULTS: We analyzed 1511 patients. Survival outcomes are in line with contemporary data. Using propensity score matching, radical and partial nephrectomy demonstrated similar perioperative and survival outcomes for clinical stage I tumors. C-indexes for ML model predictions were 0.722 for preoperative OS, 0.745 for postoperative OS, 0.682 for preoperative RFS, 0.763 for preoperative CSS, and 0.847 for postoperative CSS. For postoperative complications, sensitivity was 79.0% and positive predictive value was 75.4%. CONCLUSION: Drawing data from three decades of RCC management in a tertiary center, ML models offer reasonable predictive power, which can aid preoperative counselling, shared decision-making, and management of patient expectations before and after surgical treatment. Machine learning models drawing upon data from three decades of renal cell carcinoma (RCC) management in a tertiary center yielded robust predictive power of survival outcomes in clinical stage I RCC, which can aid preoperative counselling and management of patient expectations. 1511 patients were analyzed, and survival outcomes were in line with contemporary data. Using matched, radical and partial nephrectomy demonstrated similar perioperative and survival outcomes. Predictive accuracy of the machine learning models was favorable.
Journal IF-equivalent: 4.1 (OpenAlex 2-year mean citedness, value as of 2026-11-16, retrieved 2026-11-17; not the official Clarivate IF)Reference: Fong KY, Gan VHL, Jonathan Aslim E, Tay KJ, Chen K, Tan YG, et al. Evaluating trends in treatment and outcomes of stage I renal cell carcinoma with machine learning: The RECLAIM-L (REnal cancer landscape and artificial intelligence with machine learning) database. Ther Adv Urol. 2026 Sep 29;18:17562872261473178. doi:10.1177/17562872261473178. PMID: 43036519.Checked: Abstract only
Bladder cancer: treatment, trials and AI
- Abstract
BACKGROUND: Predicting BCG non-response in high-risk non-muscle-invasive bladder cancer (NMIBC) remains a genuine challenge in everyday urological practice. BCG engages the host immune system, and variation in individual immune responses may be reflected in measurable pre-treatment markers. OBJECTIVES: To evaluate exploratory machine learning (ML) models predicting BCG non-response at 12 months from routine pre-treatment data. DESIGN: Single-centre retrospective observational cohort study. METHODS: We reviewed 114 patients with NMIBC given intravesical BCG at one centre in Turkey, 2019-2024. By the 2021 European Association of Urology classification, all tumours were high grade and all patients high risk. BCG non-response was defined as histologically confirmed recurrence within a fixed 12-month window after at least five induction instillations, with first surveillance cystoscopy at three months. Fourteen candidate predictors, all available before the first instillation, were used; variables determined during or after treatment, including the number of BCG doses, were excluded. Penalized logistic regression, random forest and support vector machine (SVM) models were compared. All preprocessing was fitted within training folds. Performance was assessed by nested cross-validation and reported as pooled out-of-fold area under the curve (AUC) with 95% bootstrap confidence intervals. Reporting follows STROBE and TRIPOD+AI. RESULTS: Thirty-three patients (28.9%) did not respond; 18 recurrences (54.5%) were high grade. Tumour number (adjusted OR 2.701, 95% CI 1.424-5.122) and tumour size (adjusted OR 1.309 per 10 mm, 95% CI 1.012-1.693) retained an association with non-response. Random forest gave the highest pooled out-of-fold AUC (0.915, 95% CI 0.841-0.974), then SVM (0.874) and penalized logistic regression (0.703). Restricting the outcome to high-grade recurrence reduced random forest discrimination to 0.683-0.754. CONCLUSION: ML models discriminated between BCG responders and non-responders at 12 months, but discrimination was substantially lower for high-grade recurrence. Given the modest sample size and the absence of external validation, these findings should be regarded as hypothesis-generating and require prospective multicentre external validation before routine clinical implementation. Bladder cancer that has not grown into the muscle wall is usually treated by removing the tumour and then washing the bladder out with a medicine called BCG, which works by prompting the immune system to attack any remaining cancer cells. BCG helps many people, but in some patients the cancer comes back despite treatment, and at present doctors cannot tell in advance who those patients will be. We looked back at the records of 114 patients treated at a single hospital in Turkey between 2019 and 2024. All of them had aggressive tumours and all received BCG. We checked whether the cancer came back during the year after treatment finished, using the bladder examinations they had every three months. It came back in 33 of the 114 patients. We then used computer programs that learn patterns from data to see whether information already collected before treatment started — the patient's age, features of the tumour such as its size and how many there were, and results from routine blood tests — could indicate in advance who was likely to have the cancer return. We tested the programs carefully, always checking them on patients whose results they had not been shown, so that the performance we report is not simply the programs repeating what they had memorised. One of the programs separated the two groups reasonably well. However, when we looked only at the more serious returns of cancer, which are the ones that change treatment decisions, the programs performed considerably less well. Almost half of the cancers that came back were of the least aggressive type, which is normally managed by removing the tumour again rather than by more major treatment. Our study is small, comes from a single hospital and has not yet been tested anywhere else, so these programs are not ready to be used in patient care. They do suggest that routinely collected information may hold useful signals, and that this is worth testing in larger studies across several hospitals.
Journal IF-equivalent: 4.1 (OpenAlex 2-year mean citedness, value as of 2026-11-16, retrieved 2026-11-17; not the official Clarivate IF)Reference: Şığva H, Sevim M, Körpe K, Atış V, Arslan V, Görür S, et al. Machine learning–based prediction of BCG NON-RESPONSE in high-risk NON-muscle-invasive bladder cancer: An exploratory single-centre cohort study. Ther Adv Urol. 2026 Sep 28;18:17562872261493691. doi:10.1177/17562872261493691. PMID: 43036513.Checked: Abstract only
Rare cancers
Adenoid cystic carcinoma: treatment, trials and AI
No new qualifying paper for this issue.