With the incorporation of effective therapies for myelofibrosis (MF), accurately predicting outcomes after allogeneic hematopoietic cell transplantation (allo-HCT) is crucial for determining the optimal timing for this procedure. Using data from 5183 patients with MF who underwent first allo-HCT between 2005 and 2020 at European Society for Blood and Marrow Transplantation centers, we examined different machine learning (ML) models to predict overall survival after transplant. The cohort was divided into a training set (75%) and a test set (25%) for model validation. A random survival forests (RSF) model was developed based on 10 variables: patient age, comorbidity index, performance status, blood blasts, hemoglobin, leukocytes, platelets, donor type, conditioning intensity, and graft-versus-host disease prophylaxis. Its performance was compared with a 4-level Cox regression–based score and other ML-based models derived from the same data set, and with the Center for International Blood and Marrow Transplant Research score. The RSF outperformed all comparators, achieving better concordance indices across both primary and postessential thrombocythemia/polycythemia vera MF subgroups. The robustness and generalizability of the RSF model was confirmed by Akaike information criterion and time-dependent receiver operating characteristic area under the curve metrics in both sets. Although all models were prognostic for nonrelapse mortality, the RSF provided better curve separation, effectively identifying a high-risk group comprising 25% of patients. In conclusion, ML enhances risk stratification in patients with MF undergoing allo-HCT, paving the way for personalized medicine. A web application (https://gemfin.click/ebmt) based on the RSF model offers a practical tool to identify patients at high risk for poor transplantation outcomes, supporting informed treatment decisions and advancing individualized care.

Acute Myeloid Leukemia (AML) is a heterogeneous neoplasm characterized by cytogenetic and molecular alterations that drive patient prognosis. Currently established risk stratification guidelines show a moderate predictive accuracy, and newer tools that integrate multiple molecular variables have proven to provide better results. In this report, we aimed to create a new machine learning model of AML survival using gene expression data. We used gene expression data from two publicly available cohorts in order to create and validate a random forest predictor of survival, which we named ST-123. The most important variables in the model were age and the expression of KDM5B and LAPTM4B, two genes previously associated with the biology and prognostication of myeloid neoplasms. This classifier achieved high concordance indexes in the training and validation sets (0.7228 and 0.6988, respectively), and predictions were particularly accurate in patients at the highest risk of death. Additionally, ST-123 provided significant prognostic improvements in patients with high-risk mutations. Our results indicate that survival of patients with AML can be predicted to a great extent by applying machine learning tools to transcriptomic data, and that such predictions are particularly precise among patients with high-risk mutations.

Introduction: Acute myeloblastic leukemia (AML) is the most frequent type of acute leukemia in adults with an incidence of 4.2 cases per 100,000 inhabitants and poor 5-year survival. Patients with mutations in the FMS-like tyrosine kinase 3 (FLT3) gene have poor survival and higher relapse rates compared with wild-type cases. Areas covered: Several FLT3 inhibitors have been proved in FLT3^mut AML patients, with differences in their pharmacokinetics, kinase inhibitory and adverse events profiles. First-generation multi-kinase inhibitors (midostaurin, sorafenib, lestaurtinib) target multiple proteins, whereassecond-generation inhibitors (crenolanib, quizartinib, gilteritinib) are more specific and potent inhibitors of FLT3, so they are associated with less off-target toxic effects. All of these drugs have primary and acquired mechanisms of resistance, and therefore their combinations with other drugs (checkpoint inhibitors, hypomethylating agents, standard chemotherapy) and its application in different clinical settings are under study. Expert opinion: The recent clinical development of various FLT3 inhibitors for the treatment of FLT3^mut AML is an effective therapeutic strategy. However, there are unique toxicities and drug–drug interactions that need to be resolved. It is necessary to understand the mechanisms of toxicity in order to recognize and manage them adequately.

Background: FLT3 mutation is present in 25–30% of all acute myeloid leukemias (AML), and it is associated with adverse outcome. FLT3 inhibitors have shown improved survival results in AML both as upfront treatment and in relapsed/refractory disease. Curiously, a variable proportion of wild-type FLT3 patients also responded to these drugs. Methods: We analyzed 6 different transcriptomic datasets of AML cases. Differential expression between mutated and wild-type FLT3 AMLs was performed with the Wilcoxon-rank sum test. Hierarchical clustering was used to identify FLT3-mutation like AMLs. Finally, enrichment in recurrent mutations was performed with the Fisher’s test. Results: A FLT3 mutation-like gene expression pattern was identified among wild-type FLT3 AMLs. This pattern was highly enriched in NPM1 and DNMT3A mutants, and particularly in combined NPM1/DNMT3A mutants. Conclusions: We identified a FLT3 mutation-like gene expression pattern in AML which was highly enriched in NPM1 and DNMT3A mutations. Future analysis about the predictive role of this biomarker among wild-type FLT3 patients treated with FLT3 inhibitors is envisaged.

Background: Thirty to forty percent of patients with Diffuse Large B-cell Lymphoma (DLBCL) have an adverse clinical evolution. The increased understanding of DLBCL biology has shed light on the clinical evolution of this pathology, leading to the discovery of prognostic factors based on gene expression data, genomic rearrangements and mutational subgroups. Nevertheless, additional efforts are needed in order to enable survival predictions at the patient level. In this study we investigated new machine learning-based models of survival using transcriptomic and clinical data. Methods: Gene expression profiling (GEP) of in 2 different publicly available retrospective DLBCL cohorts were analyzed. Cox regression and unsupervised clustering were performed in order to identify probes associated with overall survival on the largest cohort. Random forests were created to model survival using combinations of GEP data, COO classification and clinical information. Cross-validation was used to compare model results in the training set, and Harrel’s concordance index (c-index) was used to assess model’s predictability. Results were validated in an independent test set. Results: Two hundred thirty-three and sixty-four patients were included in the training and test set, respectively. Initially we derived and validated a 4-gene expression clusterization that was independently associated with lower survival in 20% of patients. This pattern included the following genes: TNFRSF9, BIRC3, BCL2L1 and G3BP2. Thereafter, we applied machine-learning models to predict survival. A set of 102 genes was highly predictive of disease outcome, outperforming available clinical information and COO classification. The final best model integrated clinical information, COO classification, 4-gene-based clusterization and the expression levels of 50 individual genes (training set c-index, 0.8404, test set c-index, 0.7942). Conclusion: Our results indicate that DLBCL survival models based on the application of machine learning algorithms to gene expression and clinical data can largely outperform other important prognostic variables such as disease stage and COO. Head-to-head comparisons with other risk stratification models are needed to compare its usefulness.