AI and Cancer Research ── AML ── 2026-11-05
Acute myeloid leukemia (AML): treatment, trials and AI
- Abstract
BACKGROUND: Tumor lysis syndrome (TLS) is a life-threatening complication of venetoclax (VEN) for acute myeloid leukemia (AML). Existing TLS risk tools were developed for CLL/SLL and need validation in AML. This study aims to integrate drug exposure, genetic polymorphisms and clinical indicators to build a personalized machine learning model for TLS risk in AML patients on VEN, and establish a standardized management flowchart. METHODS: 186 AML patients on VEN therapy were prospectively enrolled. Key predictors were identified via univariate and multivariate regression. Four algorithms-logistic regression (LR), random forest (RF), support vector machine (SVM), and XGBoost-were compared to select the optimal predictive model. Model performance was evaluated using area under the curve (AUC), sensitivity, specificity, F1 score, calibration curves, and decision curve analysis (DCA). SHAP interpretability analysis was employed to reveal feature contributions. Finally, an online TLS risk calculator was developed, and a standardized management workflow was established. RESULTS: Among 186 patients, 39 (21.0%) developed TLS. The final model included four core covariates: VEN peak plasma concentration (Cmax), VEGFA rs3025039 genotype, ECOG performance status, elevated baseline WBC count. Among algorithms, XGBoost performed best, with validation set AUC=0.861, accuracy=0.811, specificity=0.831, sensitivity=0.7, F1=0.667, and Brier score=0.124, indicating good calibration. DCA and PRC curves confirmed high clinical utility. Based on this model, an online TLS calculator for risk stratification and personalized intervention, as well as a standardized management protocol, was successfully developed. CONCLUSION: The interpretable XGBoost model developed in this study can effectively predict TLS risk associated with VEN therapy in AML patients. The online prediction tool developed based on this model facilitates precise risk stratification, helping to avoid overtreatment of low-risk patients while optimizing resource allocation and significantly improving the safety of targeted therapy for AML.
Journal IF-equivalent: 4.7 (OpenAlex 2-year mean citedness, value as of 2026-11-03, retrieved 2026-11-05; not the official Clarivate IF)Reference: Wang C, Lin X, Yang Y, Peng M, Wu F, Zhou J, et al. Personalized Prediction of Venetoclax-Associated Tumor Lysis Syndrome in Acute Myeloid Leukemia Using Drug Exposure, Genetic Polymorphisms, and Clinical-Laboratory Characteristics. Drug Des Devel Ther. 2026 Oct;Volume 20:1-16. doi:10.2147/dddt.s627579. PMID: 42974127; PMCID: PMC13726510.Checked: Full text checked - Abstract
Acute myeloid leukemia (AML) with IDH1 mutations is amenable to targeted therapy, highlighting the need for reliable identification. We developed an attention-based multiple-instance learning model to predict IDH1 mutation status using bone marrow smear whole-slide images from 436 AML patients across four centers in Germany and the US, achieving a mean single pooled ROC-AUC of 0.81. In top tile analysis, model predictions for IDH1 were associated with immature leukemia phenotypes.
Journal IF-equivalent: 9.8 (OpenAlex 2-year mean citedness, value as of 2026-11-04, retrieved 2026-11-05; not the official Clarivate IF)Reference: Eckardt JN, Reddy CBV, Srisuwananukorn A, Pohlkamp C, Schübel M, Srivastava I, et al. Deep learning predicts IDH1 mutation status in acute myeloid leukemia from bone marrow smears. npj Precis Onc. 2026 Oct 29;10(1):412. doi:10.1038/s41698-026-01762-x.Checked: Abstract only - Abstract
Acute myeloid leukemia (AML) is a hematologic malignancy in which accurate subtype identification is important for treatment planning and prognosis. This study developed a deep learning approach for AML subtype classification using multiple single-cell images from each patient rather than conventional instance-based classification. An open-source dataset containing 81,214 single-cell images from 189 individuals was analyzed. The dataset included four genetically defined AML subtypes and a healthy control group. Two multi-view convolutional neural network (MV-CNN) models were evaluated: a 99-view model based on random sampling and a 500-view model in which image augmentation was used when fewer than 500 images were available for a patient. A single-view CNN served as the comparison model. Performance was assessed using tenfold cross-validation and accuracy, F1 score, sensitivity, specificity, precision, and Matthews correlation coefficient (MCC). The 500-view MV-CNN achieved the best performance, with an accuracy of 0.8783, F1 score of 0.8622, sensitivity of 0.8637, specificity of 0.8774, precision of 0.8616, and MCC of 0.8441. The 99-view model showed lower but comparable performance, while the single-view CNN yielded the lowest values. Patient-level multi-view analysis improved AML subtype classification compared with single-image analysis, supporting the potential of multi-view deep learning for hematologic image classification.
Journal IF-equivalent: 3.9 (OpenAlex 2-year mean citedness, value as of 2026-10-13, retrieved 2026-10-15; not the official Clarivate IF)Reference: Erdaş ÇB. Person-Based Progression in Acute Myeloid Leukemia Classification: A Multi-View Convolutional Neural Network Approach. J Imaging Inform Med. 2026 Oct 1 [Epub ahead of print]. doi:10.1007/s10278-026-02371-7. PMID: 42823569.Checked: Abstract only - Abstract
M2-type tumor-associated macrophages (TAMs) play critical roles in the immunosuppressive microenvironment and treatment resistance of acute myeloid leukemia (AML). However, their single-cell characteristics and predictive value for prognosis and therapy response remain unclear. Single-cell transcriptomic data from AML patients and healthy donors were integrated to characterize AML tumor-associated macrophages (AML-TAMs). CIBERSORT was used to evaluate M2 macrophage infiltration in the AML cohorts. WGCNA and machine learning were applied to construct an M2 macrophage-derived prognostic signature across 8 independent AML cohorts. The signature's predictive value for chemotherapy and immunotherapy response was assessed computationally, and further experimental validation was performed using clinical peripheral blood samples. Single-cell analysis revealed a distinct AML-TAM subcluster with high CD163 expression enriched in immunosuppressive and stress-adapted pathways. M2 macrophage infiltration was greater in AML patients than in controls and was associated with shorter overall survival. High M2 infiltration correlated with adverse clinical features and was negatively correlated with CD8 + T cells. An 18-gene risk score model showed robust prognostic performance across eight cohorts (log-rank P < 0.001 for each). High-risk patients exhibited chemotherapy resistance, upregulated immune checkpoint genes, and a greater predicted response to anti-PD-1 therapy. Clinical validation confirmed increased M2 abundance and key gene expression changes in AML patients. This study characterizes M2 macrophages in AML at single-cell resolution and establishes an M2 macrophage-derived signature that enables robust prediction of survival and potential therapeutic response, providing a tool for risk stratification and personalized treatment.
Journal IF-equivalent: 5.5 (OpenAlex 2-year mean citedness, value as of 2026-11-04, retrieved 2026-11-05; not the official Clarivate IF)Reference: Zhong F, Liu J, Yue J, Zhang L, Kong G. Machine learning-driven M2 macrophage signature for precision prediction of survival and therapy response in acute myeloid leukemia. Inflamm Res. 2026 Sep 27;75(1):225. doi:10.1007/s00011-026-02379-3. PMID: 42802232.Checked: Abstract only - Abstract
Despite advancements in genetic markers for acute myeloid leukemia (AML) risk stratification, outcome prediction remains challenging due to disease heterogeneity and dynamic genetic changes, highlighting the need for reliable biomarkers to improve AML treatment strategies and patient outcomes. To refine outcome predictions, we investigated the use of microbial-derived biomarkers to predict composite complete remission (CRc), relapse, and survival for patients on high- and low-intensity regimens, and to integrate those variables into the widely clinically utilized European Leukemia Network (ELN-2022) genetic risk classification model for high-intensity-treated patients. We first developed machine learning models that integrate baseline fecal metabolomics, 16S rRNA-based stool microbiome features, and clinical metadata (sex, antibiotic administration, AML somatic mutations, and cytogenetics) from two cohorts of AML patients (n = 83) undergoing remission induction chemotherapy. Univariate tests and sparse canonical correlation analysis were employed for variable selection and to explore fecal metabolite-microbe relationships. A robust machine learning approach using XGBoost was employed, with 100 stratified data splits (80% training, 20% testing) and coarse-to-fine hyperparameter optimization. Variable importance was aggregated across all models to select key predictors. For high-intensity-treated patients, XGBoost models achieved aggregated AUROC scores of 0.719, 0.729, and 0.65 for CRc, relapse, and overall survival, respectively. For low-intensity-treated patients, these models achieved aggregate AUROC scores of 0.945, 0.724, and 0.768 for these same outcomes, respectively. Integrating the biomarkers identified in the high-intensity machine-learning models with the current ELN-2022 AML risk stratification system effectively stratified patients into risk categories, which obtained higher concordance indices and likelihood ratios, demonstrating improved prognostic accuracy for each outcome compared to ELN-2022 alone. The inclusion of microbial-derived biomarkers serves as a robust prognostic tool to improve outcome prediction in AML patients, highlighting the potential of its integration into AML risk assessment and paving the way for personalized treatment strategies and improved patient outcomes.
Journal IF-equivalent: 3.9 (OpenAlex 2-year mean citedness, value as of 2026-11-04, retrieved 2026-11-05; not the official Clarivate IF)Reference: Franklin S, Sahasrabhojane P, Ivanov I, Hayase T, Hayase E, Chang C, et al. Inclusion of Multi‐Omic Biomarkers Improves Prediction Accuracy of Response, Relapse, and Overall Survival in Acute Myeloid Leukemia Patients Receiving High‐Intensity Induction Chemotherapy. Cancer Med. 2026 Sep 16;15(9):e72281. doi:10.1002/cam4.72281. PMID: 42750261; PMCID: PMC13583178.Checked: Full text checked - Abstract
Acute myeloid leukemia (AML) remains difficult to treat because profound molecular heterogeneity enables rapid adaptive resistance to targeted therapy. We developed an AI-driven pharmacogenomic pipeline that integrates genomic, transcriptomic, epigenetic, and clinical features to predict patient-specific response to venetoclax and translate resistance-associated signals into actionable therapeutic hypotheses. Across internal and external cohorts, the deep learning model showed strong discrimination and generalization (AUROC 0.84-0.90), produced calibrated probabilities, and stratified overall survival independent of standard prognostic factors. Explainable modeling (SHAP) identified mechanistically coherent drivers of sensitivity and resistance: high BCL2 and apoptotic priming favored response, whereas MCL1 upregulation, TP53 disruption, and RAS/MAPK activation were dominant resistance programs. Translating these findings into therapy design, network-based modeling and in silico perturbation prioritized rational combinations expected to block escape routes, including venetoclax plus MCL1 inhibition (e.g., AZD5991-class inhibitors), venetoclax plus FLT3 inhibition (e.g., gilteritinib-class agents) in signaling-driven disease, and venetoclax plus p53-axis modulation (e.g., MDM2 inhibition) in TP53-altered contexts. Structural candidate evaluation using ensemble docking provided supportive drug-target interaction evidence for prioritized dependencies. Together, these results establish a clinically interpretable, resistance-aware AI framework for precision optimization of venetoclax-based combination therapy in AML.
Journal IF-equivalent: 9.8 (OpenAlex 2-year mean citedness, value as of 2026-11-04, retrieved 2026-11-05; not the official Clarivate IF)Reference: Wang C, You Z, Chen S, He Y, Chen X, Qu S, et al. AI-enabled multi-omics pharmacogenomic modeling guides resistance-aware multitarget optimization of venetoclax therapy in acute myeloid leukemia. NPJ Precis Oncol. 2026 Sep 10;10(1):358. doi:10.1038/s41698-026-01608-6. PMID: 42778648; PMCID: PMC13601629.Checked: Full text checked - Abstract
Feline McDonough sarcoma (FMS)-like tyrosine kinase 3 internal tandem duplication (FLT3-ITD) mutations are clinically important in acute myeloid leukaemia (AML). We developed and validated a machine learning model using routine laboratory tests to estimate the probability of FLT3-ITD mutations in newly diagnosed AML to prioritise rapid molecular testing. Consecutive patients treated at Ruijin Hospital during 2021-2024 were included. Those diagnosed in 2021-2023 comprised the training cohort, and those in 2024 formed the internal temporal validation cohort, and the BeatAML2 cohort served as external validation. Following Boruta feature selection, seven machine learning algorithms were evaluated via area under the receiver operating characteristic curve (AUC). The optimal model was interpreted using SHapley Additive exPlanations (SHAP) and deployed as a web-based calculator. Overall, 514 patients were included (training, n = 254; validation, n = 84; external validation, n = 176). Boruta algorithm selected five candidate features, and a simple four-feature model was constructed. Gradient boosting machine (GBM) performed best, with AUCs of 0.847, 0.772 and 0.760 in the training, internal validation and external validation cohorts respectively. SHAP analysis showed absolute blast count contributed most to the model. The GBM model may help identify patients with newly diagnosed AML who should be prioritised for rapid molecular testing for FLT3-ITD mutations.
Journal IF-equivalent: 3.0 (OpenAlex 2-year mean citedness, value as of 2026-11-04, retrieved 2026-11-05; not the official Clarivate IF)Reference: Song J, Gu Z, Han J, Zhang Y, Zhao L, Yan H, et al. Development and validation of a machine learning model based on routine laboratory tests for predicting FLT3 ‐ITD mutations in acute myeloid leukaemia. Br J Haematol. 2026 Sep 2 [Epub ahead of print]. doi:10.1111/bjh.70829. PMID: 42686039.Checked: Abstract only - Abstract
[Figure: see text]
Journal IF-equivalent: 7.4 (OpenAlex 2-year mean citedness, value as of 2026-11-04, retrieved 2026-11-05; not the official Clarivate IF)Reference: Yates SJ, Terman E, Miller L, Cursio JF, Wild C, Huisingh-Scheetz M, et al. Association of sarcopenia with treatment tolerance in older adults with acute myeloid leukemia using artificial intelligence computed tomography–segmentation and a geriatric assessment. Leukemia. 2026 Aug 31;40(11):2611-5. doi:10.1038/s41375-026-03099-7. PMID: 42675142; PMCID: PMC13713559.Checked: Full text checked - Abstract
Allogeneic hematopoietic stem cell transplantation (allo-HSCT) is a curative treatment option for a significant proportion of patients with acute myeloid leukemia (AML), and it is generally recommended when the relapse risk without allo-HSCT outweighs the estimated non-relapse mortality significantly. While current recommendations for allo-HSCT are based on risk groups, there is considerable heterogeneity within these individual categories. We analyzed 2550 intensively treated AML patients aged 18-70 with cytogenetic and next-generation sequencing data from the HARMONY Alliance database, who did not receive allo-HSCT in first complete remission (CR1). A non-parametric machine learning (ML) model based on Bayesian Additive Regression Trees (BART) integrated clinical variables and genomic aberrations to provide individualized outcome estimations. External validation was performed in a cohort of 714 patients enrolled in UK-NCRI trials. The predictive performance of the HARMONY ML model, measured by the area under the time-dependent receiver operating curve (AUC(t)), was superior to European LeukemiaNet (ELN)2022 risk classification in estimating 5-year overall survival (0.741 vs. 0.700), 5-year relapse-free survival (0.752 vs. 0.705), and 5-year cumulative incidence of relapse (0.742 vs. 0.708), which was confirmed in the external validation cohort. Notably, the model revealed substantial heterogeneity within ELN2022 risk groups, identifying a significant proportion of favorable-risk patients with a predicted 5-year CIR > 40%, who could potentially benefit from allo-HSCT in CR1. The HARMONY ML model provides individualized risk prediction in intensively treated adult AML patients and supports more tailored therapeutic decisions regarding allo-HSCT in CR1, which should be further advanced by integrating measurable residual disease in the future.
Journal IF-equivalent: 3.7 (OpenAlex 2-year mean citedness, value as of 2026-11-04, retrieved 2026-11-05; not the official Clarivate IF)Reference: Hernández‐Sánchez A, Martínez Elicegui J, Sträng E, Benner A, Sobas M, Villaverde Ramiro Á, et al. Machine learning provides individualized prediction of outcomes after first complete remission without allo‐HSCT consolidation in adult acute myeloid leukemia—A HARMONY study. Hemasphere. 2026 Aug 30;10(9):e70461. doi:10.1002/hem3.70461. PMID: 42670491; PMCID: PMC13526294.Checked: Full text checked - Abstract
Cyclin-dependent kinase 8 (CDK8) has emerged as a promising therapeutic target for acute myeloid leukaemia (AML). We developed an integrated computational pipeline combining neural network-based potency prediction with molecular dynamics (MD) simulations for CDK8 inhibitor discovery. A curated dataset of 1,200 unique CDK8 inhibitors was assembled from ChEMBL. The optimal neural network architecture (two hidden layers, 512→128 units) with dropout regularization achieved a test set r 2 of 0.47 and RMSE of 0.87. Y-randomization testing confirmed genuine structure-activity relationships. Using an extrapolation-focused training strategy combined with a SELFIES-based genetic recombination algorithm, we generated novel molecular structures beyond the training distribution. SHAP analysis revealed fingerprint bits as critical determinants of potency. Applicability‑domain analysis confirmed that the novel hit falls within validated chemical space. The identified candidate exhibited a predicted pIC 50 of 11.97, substantially exceeding the most potent training compound (pIC 50 = 10.09) and compound 12 (pIC 50 = 7.47). Molecular docking revealed Moldock scores of -143.9 kcal/mol for the novel hit versus -113.9 kcal/mol for compound 12. MD simulations demonstrated stable binding with the novel hit forming a highly stable hydrogen bond with Asp98. MM-GBSA calculations showed superior binding free energy for the novel hit (-99.10 vs. -47.77 kcal/mol).
Journal IF-equivalent: 2.9 (OpenAlex 2-year mean citedness, value as of 2026-11-04, retrieved 2026-11-05; not the official Clarivate IF)Reference: Awad BR, Sargolzaei M, Nikoofard H. An integrated deep learning and molecular dynamics pipeline for the discovery of novel CDK8 inhibitors against acute myeloid leukaemia. SAR QSAR Environ Res. 2026 Aug 7;37(7):619-53. doi:10.1080/1062936x.2026.2709353. PMID: 42565435.Checked: Abstract only - Abstract
Current platforms for drug screening typically do not account for the tumor microenvironment. Microfluidic shear flow assays provide a highly sensitive tool for studying tumor cell communication at the single cell level with the microenvironment. Adhesion thereby serves as a functional readout that reflects cellular state, including loss of viability. However, previous platforms required extensive manual handling and time-consuming post-assay analysis. We developed a bright-field microscopy-enabled, semi-automated shear flow platform that combines hardware operation with a machine-learning-based analysis pipeline. The algorithm delivers consistent, high-quality results within minutes with a precision of 98.3% and a recall of 99.1%, indicative of high tracking specificity and object discrimination.
Journal IF-equivalent: 3.5 (OpenAlex 2-year mean citedness, value as of 2026-11-04, retrieved 2026-11-05; not the official Clarivate IF)Reference: Ashok D, Raith D, Kalweit G, Kalweit M, Klett A, Duyster J, et al. A Semi‐Automated Microfluidic Platform Employing Machine Learning Analysis to Study Adhesion Kinetics in Acute Myeloid Leukemia. Microsc Res Tech. 2026 Jul 22 [Epub ahead of print]. doi:10.1002/jemt.70161. PMID: 42487303.Checked: Abstract only - Abstract
Lysosomes are essential for intracellular degradation and recycling, and changes in their function significantly contribute to tumor growth. Nonetheless, the exact role of lysosome-related genes (LRGs) in the pathogenesis of acute myeloid leukemia (AML) is still inadequately comprehended. Differentially expressed LRGs (DE-LRGs) between AML and control groups were identified using AML-related data extracted from the Gene Expression Omnibus (GEO). The LRGs-related prognostic genes were identified and the risk model was established using univariate COX regression analysis and machine learning algorithms, based on the data obtained from The Cancer Genome Atlas (TCGA). Subsequently, we performed comprehensive analyses regarding clinical features, functional pathways, immune microenvironment, and chemotherapeutic drugs sensitivity between the high- and low-risk groups. Reverse transcription Quantitative polymerase chain reaction (RT-qPCR) and western blot were adopted to validate the expression of prognostic genes in human bone marrow-derived cell line HS-27 A and human AML cell line MOLM-13. Through comprehensive analysis, a risk model was developed utilizing ten LRGs (ATP6V0E2, CALCRL, TMEM165, GZMB, HCK, TCIRG1, CD1D, GPRASP1, ABCA1, and NAGA), and this model was further validated using GEO datasets. Significant differences in clinical characteristics, functional pathways, immune microenvironment characteristics, and chemotherapeutic drug sensitivity were observed between the two risk groups In vitro validation experiment illustrated that the expression trends of ATP6V0E2, TMEM165, and ABCA1 were consistent with our bioinformatics analysis. Our study demonstrates that lysosome-associated signature might forecast the prognosis of AML patients and offer guidance for subsequent immunotherapy and chemotherapy strategies.
Journal IF-equivalent: 2.4 (OpenAlex 2-year mean citedness, value as of 2026-10-12, retrieved 2026-10-13; not the official Clarivate IF)Reference: Lu L, Jiang X, Lu J, Lv J. Machine learning-based integration develops a novel lysosome-related prognostic signature associated with prognosis and immune infiltration landscape in acute myeloid leukemia. Discov Oncol. 2026 Jun 23;17(1):1308. doi:10.1007/s12672-026-05463-6. PMID: 42334654; PMCID: PMC13550311.Checked: Full text checked - Abstract
Elderly patients with acute myeloid leukemia (AML) exhibit considerable biological and clinical heterogeneity, hindering precise prognosis. Existing prognostic systems inadequately capture the complexity of elderly AML due to their reliance on data from younger cohorts and omission of key factors like immunophenotypic markers and therapeutic profiles. This study aimed to develop and internally validate a machine learning-based prognostic model specifically tailored to elderly AML patients. A total of 156 patients were analyzed using a two-stage modeling strategy. Clinical and genomic variables were modeled first, followed by independent analysis of immunophenotypic features. Feature selection was performed using multilayer perceptron (MLP) and random forest (RF), while multivariate Cox regression was used for final model construction. Internal validation was conducted using 1000 bootstrap iterations to assess model stability and performance. The model demonstrated strong predictive performance, with a concordance index (C-index) of 0.702. Time-dependent area under the curve (AUC) and calibration plots confirmed accurate prediction of 1-, 3-, and 5-year overall survival. Decision curve analysis indicated favorable net benefit across a range of threshold probabilities. Key independent prognostic factors identified included TP53 mutations, high CD13 expression, and IDH2 mutations. This model provides a robust and interpretable tool for individualized risk stratification in elderly AML. By integrating genomic, immunophenotypic, and therapeutic variables, it may help optimize treatment decisions and improve outcomes for this vulnerable population. Future efforts should focus on external validation and integration of dynamic biomarkers.
Journal IF-equivalent: 4.6 (OpenAlex 2-year mean citedness, value as of 2026-11-04, retrieved 2026-11-05; not the official Clarivate IF)Reference: Zhang L, Liu J, Liang J, Zhang X, Xin J, Wei M, et al. Machine learning-guided risk stratification in elderly AML based on genomic, immunophenotypic and therapeutic profiles. BMC Geriatr. 2026 Jun 22;26(1):1124. doi:10.1186/s12877-026-07734-x. PMID: 42324513.Checked: Abstract only - Abstract
Background/Objectives: To date, no validated scoring system can accurately predict the responses of acute myeloid leukaemia (AML) patients to induction chemotherapy (CTX). Current risk assessment relies on complex cytogenetic and molecular abnormalities and focuses on mutations in genes considered fundamental to leukaemogenesis.
Methods: We performed bioinformatic analysis of targeted sequencing (TS) data from 111 genes in 1552 AML patients, focusing on mutational patterns derived from single-nucleotide variant (SNV) catalogues. The SNV catalogues were analysed using non-negative matrix factorisation (NNMF), a linear dimensionality-reduction approach, to extract risk-defining recursive signatures (RSs) and to distinguish responders from resistant patients following induction CTX. To enable patient-level prediction, we complemented NNMF with a Random Forest (RF) model. Given the class imbalance between responders and resistant cases, model performance was improved by applying the Synthetic Minority Over-sampling Technique (SMOTE) and by incorporating germline variants alongside somatic mutations.
Results: NNMF-derived RSs captured clinically relevant structures in patients' mutational profiles and clustered patients by treatment response, indicating that the diagnostic targeted sequencing data contain sufficient information for risk stratification and treatment response prediction. At the single-patient level, RF models incorporating balanced data and germline variation improved predictive performance compared with unbalanced somatic-only models.
Conclusions: These findings demonstrate that machine learning applied to targeted sequencing data can extract clinically informative mutational structures and improve risk stratification in AML, supporting its potential integration into precision treatment decision-making.
Journal IF-equivalent: 5.3 (OpenAlex 2-year mean citedness, value as of 2026-10-07, retrieved 2026-10-08; not the official Clarivate IF)Reference: Elhaddad H, Chiriches C, Nandi SP, van Eijk P, Gilkes A, Watts K, et al. Mutational Signatures and Machine Learning for Risk Stratification of Acute Myeloid Leukaemia Based on Targeted Sequencing Data. Cancers (Basel). 2026 Jun 12;18(12):1925. doi:10.3390/cancers18121925. PMID: 42352459.Checked: Abstract only - Abstract
Acute Myeloid Leukemia (AML) is a heterogeneous hematological malignancy with poor prognosis, despite therapeutic advances. Gene expression analysis has emerged as a powerful tool for identifying novel biological markers that can aid in predicting patient outcomes. This study is aimed to identify prognostic gene expression signatures from RNA-seq data of 457 AML patients. Filtering methods were applied to reduce the set of genes, resulting in a subset of 685. Lasso-Cox, Bayesian Model Averaging (BMA), Random Survival Forest (RSF), and a Cox-time Neural Network (CtNN) were implemented to select the most promising prognostic genes. Explainable tools such as SHapley Additive exPlanations (SHAP) and Local Interpretable Model-agnostic Explanations (LIME) were used to gain a deeper understanding of neural network outputs. An enrichment analysis was performed to identify enriched biological pathways. SHAP and LIME interpretations of the CtNN output identified gene sets that differed from those selected by the other algorithms. According to the two-validation metrics the SHAP-derived signature demonstrated superior performance in the testing set. (C-index[95%CI]: 0.639[0.581-0.696] and Integrated Brier Score [95%CI]: 0.172[0.161-0.183]). Enrichment analysis revealed structural and developmental pathways, emphasizing the role of microtubule dynamics and ciliary-related signaling in the bone marrow microenvironment.
Journal IF-equivalent: 2.7 (OpenAlex 2-year mean citedness, value as of 2026-11-04, retrieved 2026-11-05; not the official Clarivate IF)Reference: Blanda A, Manitto R, Pizzamiglio S, Sangaletti S, Verderio P. Harnessing cutting-edge techniques to identify novel gene expression signatures in acute myeloid leukemia patients. Front Bioinform. 2026 Jun 8;6:1805826. doi:10.3389/fbinf.2026.1805826. PMID: 42339130; PMCID: PMC13284140.Checked: Full text checked - Abstract
We developed machine-learning models to predict isocitrate dehydrogenase (IDH) mutation status in acute myeloid leukemia (AML) from gene expression profiles and to reconstruct missing IDH annotations across public datasets. Transcriptomic data from 19 cohorts (5844 samples) were harmonized using batch correction, and 1546 samples with known IDH status were used to train a feed-forward neural network and a logistic regression (LR) classifier within a nested cross-validation framework, followed by independent validation in the TCGA-LAML dataset. The LR model showed superior performance, achieving receiver operating characteristic area under the curve [Formula: see text] 0.994 [Formula: see text] 0.007, accuracy [Formula: see text] 0.983 [Formula: see text] 0.006, balanced accuracy [Formula: see text] 0.979 [Formula: see text] 0.005, sensitivity for the IDH-mutant (IDH-MUT) class [Formula: see text] 0.972 [Formula: see text] 0.010, and specificity [Formula: see text] 0.986 [Formula: see text] 0.008, and correctly classified all IDH-MUT cases in the independent cohort. Applying the final model to samples lacking annotations enabled reconstruction of IDH status for 4148 AML cases, expanding the number of molecularly characterized transcriptomes available for downstream analyses. Predicted groups recapitulated known IDH-associated transcriptional signatures, supporting biological validity. This work demonstrates that IDH mutation status can be accurately inferred from transcriptomic data alone and provides a scalable framework to recover missing genomic annotations, thereby enhancing the utility of public AML resources for large-scale biological and translational research.
Journal IF-equivalent: 4.7 (OpenAlex 2-year mean citedness, value as of 2026-11-04, retrieved 2026-11-05; not the official Clarivate IF)Reference: Jung I, Vitte AL, Chuffart F, Chevalier S, Mossuz P, Khochbin S, et al. Predicting isocitrate dehydrogenase mutation status in acute myeloid leukemia from gene expression profiles by machine learning. NAR Genom Bioinform. 2026 Jun 8;8(2):lqag058. doi:10.1093/nargab/lqag058. PMID: 42266574; PMCID: PMC13244157.Checked: Full text checked - Abstract
Background: Acute myeloid leukemia (AML) remains a lethal hematologic malignancy with high heterogeneity. Macrophage-mediated efferocytosis in the tumor microenvironment is implicated in immune suppression and disease progression.
Methods: We integrated single-cell and bulk transcriptomic data from public cohorts to identify genes associated with macrophages and efferocytosis in AML. Candidate genes were screened for prognosis using univariate Cox regression. A comprehensive machine learning framework, evaluating 117 algorithm combinations, was employed to construct a robust prognostic model. The optimal LASSO and random survival forest approach identified CD52 and S100A4 as core prognostic genes. The resulting two-gene model was rigorously validated using Kaplan-Meier analysis, time-dependent ROC curves, and calibration plots across multiple independent cohorts. The associations of the risk score with the immune microenvironment and drug sensitivity were further analyzed. SHapley Additive exPlanations (SHAP) analysis was applied to interpret the model's decision-making.
Results: The two-gene signature demonstrated stable and powerful predictive performance for overall survival in both training and external validation sets. The risk score was an independent prognostic factor and showed significant correlations with immune cell infiltration patterns and response to chemotherapeutic agents. SHAP analysis confirmed the consistent and biologically plausible contributions of CD52 and S100A4. Single-cell resolution analysis revealed their specific enrichment in distinct AML-associated macrophage subpopulations.
Conclusions: We developed a novel macrophage efferocytosis-based prognostic model using a multi-omics and machine learning approach. This model provides valuable insights into the immune microenvironment of AML and offers a potential tool for risk stratification and therapeutic guidance.
Journal IF-equivalent: 2.4 (OpenAlex 2-year mean citedness, value as of 2026-11-04, retrieved 2026-11-05; not the official Clarivate IF)Reference: Li X, Zhang Y, Wei X, Huang R. A macrophage-related efferocytosis-based two-gene prognostic model for acute myeloid leukemia identified by multi-omics and machine learning. Ann Hematol. 2026 May 29;105(8):363. doi:10.1007/s00277-026-07044-7. PMID: 42213195; PMCID: PMC13447400.Checked: Full text checked - Abstract
Missing data is a challenge in clinical research, especially in real-world data (RWD), where complete case analysis can bias results and reduce power. Ensemble learning approaches like Super Learner (SL) show strong numerical performance for prediction problems, but their use for missing value imputation (MVI) in oncology datasets is unexplored. We sought to develop and evaluate a novel SL-based imputation function that can impute multiple variables and quantify observation-specific uncertainty. We analyzed two independent cohorts of acute myeloid leukemia patients (n = 1641), 546 patients from the University of Colorado and 1095 from an external real-world cohort. The SL-based MVI function includes data processing, predictor selection, binary and continuous variable pipelines, and automatic performance measurement. Ensembles for both binary and continuous variables integrate diverse base learners, such as generalized linear models, random forests, and neural networks, via a meta learner that optimizes predictive accuracy. The binary variable pipeline's SL ensemble was optimized using area under the curve (AUC), while the continuous variable pipeline's SL ensemble was optimized via non-negative least squares. Performance was compared to multiple imputation by chained equations (MICE) using balanced accuracy, F 1 -score, root mean square error (RMSE), and visualizations. Observation-specific uncertainty was quantified for all imputations of both binary and continuous variables, with both additionally having lower and upper resampling-based potential imputation values. The SL cross-validation loop, SL ensemble trained for imputation, and resampling all supported parallelization. Clinically significant features of the cohorts were selected a priori based on prior literature. In a numerical experiment with 9 clinically important binary features, the proposed MVI function imputed and achieved higher balanced accuracy than MICE for 7/9 variables (mean balanced accuracy 89.04% vs. 80.75%) with comparable performance for the other 2 variables. The continuous variable SL ensemble, across 4 variables, showed an average 24.45% lower RMSE than MICE. On average, the SL ensemble trained for prediction took 145.02 s to process for binary targets. This study demonstrates that the SL-based imputation function has improved performance over MICE in high-dimensional RWD while providing novel, observation-level uncertainty quantification.
Journal IF-equivalent: 6.2 (OpenAlex 2-year mean citedness, value as of 2026-11-04, retrieved 2026-11-05; not the official Clarivate IF)Reference: Zbysinski T, Wu L, Dale J, Coates J, Sapiah K, Reuben J, et al. A comparative analysis in a clinical cohort: multiple imputation by chained equations and a novel super learner-based imputation approach. BMC Med Res Methodol. 2026 May 12;26(1):148. doi:10.1186/s12874-026-02858-5. PMID: 42121053; PMCID: PMC13339447.Checked: Full text checked