Refining risk prediction in pediatric acute lymphoblastic leukemia through DNA methylation profiling.

Acute lymphoblastic leukemia (ALL) is the most prevalent cancer in children, and despite considerable progress in treatment outcomes, relapses still pose significant risks of mortality and long-term complications. To address this challenge, we employed a supervised machine learning technique, specifically random survival forests, to predict the risk of relapse and mortality using array-based DNA methylation data from a cohort of 763 pediatric ALL patients treated in Nordic countries. The relapse risk predictor (RRP) was constructed based on 16 CpG sites, demonstrating c-indexes of 0.667 and 0.677 in the training and test sets, respectively. The mortality risk predictor (MRP), comprising 53 CpG sites, exhibited c-indexes of 0.751 and 0.754 in the training and test sets, respectively. To validate the prognostic value of the predictors, we further analyzed two independent cohorts of Canadian (n = 42) and Nordic (n = 384) ALL patients. The external validation confirmed our findings, with the RRP achieving a c-index of 0.667 in the Canadian cohort, and the RRP and MRP achieving c-indexes of 0.529 and 0.621, respectively, in an independent Nordic cohort. The precision of the RRP and MRP models improved when incorporating traditional risk group data, underscoring the potential for synergistic integration of clinical prognostic factors. The MRP model also enabled the definition of a risk group with high rates of relapse and mortality. Our results demonstrate the potential of DNA methylation as a prognostic factor and a tool to refine risk stratification in pediatric ALL. This may lead to personalized treatment strategies based on epigenetic profiling.

Validation of the Artificial Intelligence Prognostic Scoring System for Myelodysplastic Syndromes in chronic myelomonocytic leukaemia: A novel approach for improved risk stratification.

Chronic myelomonocytic leukaemia (CMML) is a rare haematological disorder characterized by monocytosis and dysplastic changes in myeloid cell lineages. Accurate risk stratification is essential for guiding treatment decisions and assessing prognosis. This study aimed to validate the Artificial Intelligence Prognostic Scoring System for Myelodysplastic Syndromes (AIPSS-MDS) in CMML and to assess its performance compared with traditional scores using data from a Spanish registry (n = 1343) and a Taiwanese hospital (n = 75). In the Spanish cohort, the AIPSS-MDS accurately predicted overall survival (OS) and leukaemia-free survival (LFS), outperforming the Revised-IPSS score. Similarly, in the Taiwanese cohort, the AIPSS-MDS demonstrated accurate predictions for OS and LFS, showing superiority over the IPSS score and performing better than the CPSS and molecular CPSS scores in differentiating patient outcomes. The consistent performance of the AIPSS-MDS across both cohorts highlights its generalizability. Its adoption as a valuable tool for personalized treatment decision-making in CMML enables clinicians to identify high-risk patients who may benefit from different therapeutic interventions. Future studies should explore the integration of genetic information into the AIPSS-MDS to further refine risk stratification in CMML and improve patient outcomes.

Incidence, risk factors, and outcomes of second neoplasms in patients with acute promyelocytic leukemia: the PETHEMA-PALG experience.

The most important challenges in acute promyelocytic leukemia (APL) is preventing early death and reducing long-term events, such as second neoplasms (s-NPLs). We performed a retrospective analysis of 2670 unselected APL patients, treated with PETHEMA "chemotherapy based" and "chemotherapy free" protocols. Only de novo APL patients who achieved complete remission (CR) and completed the three consolidation cycles were enrolled into the analysis. Out of 2670 APL patients, there were 118 (4.4%) who developed s-NPLs with the median latency period (between first CR and diagnosis of s-NPL) of 48.0 months (range 2.8-231.1): 43.3 (range: 2.8-113.9) for s-MDS/AML and 61.7 (range: 7.1-231.1) for solid tumour. The 5-year CI of all s-NPLs was of 4.43% and 10 years of 7.92%. Among s-NPLs, there were 58 cases of s-MDS/AML, 3 cases of other hematological neoplasms, 57 solid tumours and 1 non-identified neoplasm. The most frequent solid tumour was colorectal, lung and breast cancer. Overall, the 2-year OS from diagnosis of s-NPLs was 40.6%, with a median OS of 11.1 months. Multivariate analysis identified age of 35 years (hazard ratio = 0.2584; p < 0.0001) as an independent prognostic factor for s-NPLs. There were no significant differences in CI of s-NPLs at 5 years between chemotherapy-based vs chemotherapy-free regimens (hazard ratio = 1.09; p = 0.932). Larger series with longer follow-up are required to confirm the potential impact of ATO+ATRA regimens to reduce the incidence of s-NPLs after front-line therapy for APL.

Machine Learning Improves Risk Stratification in Myelodysplastic Neoplasms: An Analysis of the Spanish Group of Myelodysplastic Syndromes.

Myelodysplastic neoplasms (MDS) are a heterogeneous group of hematological stem cell disorders characterized by dysplasia, cytopenias, and increased risk of acute leukemia. As prognosis differs widely between patients, and treatment options vary from observation to allogeneic stem cell transplantation, accurate and precise disease risk prognostication is critical for decision making. With this aim, we retrieved registry data from MDS patients from 90 Spanish institutions. A total of 7202 patients were included, which were divided into a training (80%) and a test (20%) set. A machine learning technique (random survival forests) was used to model overall survival (OS) and leukemia-free survival (LFS). The optimal model was based on 8 variables (age, gender, hemoglobin, leukocyte count, platelet count, neutrophil percentage, bone marrow blast, and cytogenetic risk group). This model achieved high accuracy in predicting OS (c-indexes; 0.759 and 0.776) and LFS (c-indexes; 0.812 and 0.845). Importantly, the model was superior to the revised International Prognostic Scoring System (IPSS-R) and the age-adjusted IPSS-R. This difference persisted in different age ranges and in all evaluated disease subgroups. Finally, we validated our results in an external cohort, confirming the superiority of the Artificial Intelligence Prognostic Scoring System for MDS (AIPSS-MDS) over the IPSS-R, and achieving a similar performance as the molecular IPSS. In conclusion, the AIPSS-MDS score is a new prognostic model based exclusively on traditional clinical, hematological, and cytogenetic variables. AIPSS-MDS has a high prognostic accuracy in predicting survival in MDS patients, outperforming other well-established risk-scoring systems.

Corrigendum: Evaluation of the Stellae-123 prognostic gene expression signature in acute myeloid leukemia.

[This corrects the article DOI: 10.3389/fonc.2022.968340.].

Presence of Myeloid Mutations in Patients with Chronic Myeloid Leukemia Increases Risk of Cardiovascular Event on Tyrosine Kinase Inhibitor Treatment.

For chronic myeloid leukemia (CML) patients with a known risk of cardiovascular events (CVE), imatinib is often recommended for first-line tyrosine kinase inhibitor (TKI) treatment rather than a second-generation TKI (2G-TKI) such as nilotinib or dasatinib. To date, very few studies have evaluated the genetic predisposition associated with CVE development on TKI treatment. In this retrospective study of 102 CML patients, 26 CVEs were reported during an average follow-up of over 10 years. Next-generation sequencing identified pathogenic/likely pathogenic mutations in genes associated with myeloid malignancies in 24.5% of the diagnostic samples analyzed. Patients with a recorded CVE had more myeloid mutations (0.48 vs. 0.14, p = 0.019) and were older (65.1 vs. 55.7 years, p = 0.016). Age ≥ 60 years and receiving a 2G-TKI in first-line were CVE risk factors. The presence of a pathogenic somatic myeloid mutation was an independent risk factor for CVE on any TKI (HR 2.79, p = 0.01), and significantly shortened the CV event-free survival of patients who received first-line imatinib (by 70 months, p = 0.011). Indeed, 62% of patients on imatinib with mutations had a CVE vs. the 19% on imatinib with a mutation and no CVE. In conclusion, myeloid mutations detectable at diagnosis increase CVE risk, particularly for patients on imatinib, and might be considered for first-line TKI choice.

A prognostic model based on gene expression parameters predicts a better response to bortezomib-containing immunochemotherapy in diffuse large B-cell lymphoma.

Diffuse Large B-cell Lymphoma (DLBCL) is the most common type of aggressive lymphoma. Approximately 60% of fit patients achieve curation with immunochemotherapy, but the remaining patients relapse or have refractory disease, which predicts a short survival. Traditionally, risk stratification in DLBCL has been based on scores that combine clinical variables. Other methodologies have been developed based on the identification of novel molecular features, such as mutational profiles and gene expression signatures. Recently, we developed the LymForest-25 profile, which provides a personalized survival risk prediction based on the integration of transcriptomic and clinical features using an artificial intelligence system. In the present report, we studied the relationship between the molecular variables included in LymForest-25 in the context of the data released by the REMoDL-B trial, which evaluated the addition of bortezomib to the standard treatment (R-CHOP) in the upfront setting of DLBCL. For this, we retrained the machine learning model of survival on the group of patients treated with R-CHOP (N=469) and then made survival predictions for those patients treated with bortezomib plus R-CHOP (N=459). According to these results, the RB-CHOP scheme achieved a 30% reduction in the risk of progression or death for the 50% of DLBCL patients at higher molecular risk (p-value 0.03), potentially expanding the effectiveness of this treatment to a wider patient population as compared with other previously defined risk groups.

Evaluation of the Stellae-123 prognostic gene expression signature in acute myeloid leukemia.

Risk stratification in acute myeloid leukemia (AML) has been extensively improved thanks to the incorporation of recurrent cytogenomic alterations into risk stratification guidelines. However, mortality rates among fit patients assigned to low or intermediate risk groups are still high. Therefore, significant room exists for the improvement of AML prognostication. In a previous work, we presented the Stellae-123 gene expression signature, which achieved a high accuracy in the prognostication of adult patients with AML. Stellae-123 was particularly accurate to restratify patients bearing high-risk mutations, such as ASXL1, RUNX1 and TP53. The intention of the present work was to evaluate the prognostic performance of Stellae-123 in external cohorts using RNAseq technology. For this, we evaluated the signature in 3 different AML cohorts (2 adult and 1 pediatric). Our results indicate that the prognostic performance of the Stellae-123 signature is reproducible in the 3 cohorts of patients. Additionally, we evidenced that the signature was superior to the European LeukemiaNet 2017 and the pediatric clinical risk scores in the prediction of survival at most of the evaluated time points. Furthermore, integration with age substantially enhanced the accuracy of the model. In conclusion, Stellae-123 is a reproducible machine learning algorithm based on a gene expression signature with promising utility in the field of AML.

Prognostic Stratification of Diffuse Large B-cell Lymphoma Using Clinico-genomic Models: Validation and Improvement of the LymForest-25 Model.

Diffuse large B-cell lymphoma (DLBCL) is the most common type of non-Hodgkin lymphoma. Despite notable therapeutic advances in the last decades, 30%-40% of affected patients develop relapsed or refractory disease that frequently precludes an infamous outcome. With the advent of new therapeutic options, it becomes necessary to predict responses to the standard treatment based on rituximab, cyclophosphamide, doxorubicin, vincristine, and prednisone (R-CHOP). In a recent communication, we presented a new machine learning model (LymForest-25) that was based on 25 clinical, biochemical, and gene expression variables. LymForest-25 achieved high survival prediction accuracy in patients with DLBCL treated with upfront immunochemotherapy. In this study, we aimed to evaluate the performance of the different features that compose LymForest-25 in a new UK-based cohort, which contained 481 patients treated with upfront R-CHOP for whom clinical, biochemical and gene expression information for 17 out of 19 transcripts were available. Additionally, we explored potential improvements based on the integration of other gene expression signatures and mutational clusters. The validity of the LymForest-25 gene expression signature was confirmed, and indeed it achieved a substantially greater precision in the estimation of mortality at 6 months and 1, 2, and 5 years compared with the cell-of-origin (COO) plus molecular high-grade (MHG) classification. Indeed, this signature was predictive of survival within the MHG and all COO subgroups, with a particularly high accuracy in the "unclassified" group. Integration of this signature with the International Prognostic Index (IPI) score provided the best survival predictions. However, the increased performance of molecular models with the IPI score was almost exclusively restricted to younger patients (<70 y). Finally, we observed a tendency towards an improved performance by combining LymForest-25 with the LymphGen mutation-based classification. In summary, we have validated the predictive capacity of LymForest-25 and expanded the potential for improvement with mutation-based prognostic classifications.

Safety of FLT3 inhibitors in patients with acute myeloid leukemia.

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 FLT3mut 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 FLT3mut 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.

Survival prediction and treatment optimization of multiple myeloma patients using machine-learning models based on clinical and gene expression data.

Multiple myeloma (MM) remains mostly an incurable disease with a heterogeneous clinical evolution. Despite the availability of several prognostic scores, substantial room for improvement still exists. Promising results have been obtained by integrating clinical and biochemical data with gene expression profiling (GEP). In this report, we applied machine learning algorithms to MM clinical and RNAseq data collected by the CoMMpass consortium. We created a 50-variable random forests model (IAC-50) that could predict overall survival with high concordance between both training and validation sets (c-indexes, 0.818 and 0.780). This model included the following covariates: patient age, ISS stage, serum B2-microglobulin, first-line treatment, and the expression of 46 genes. Survival predictions for each patient considering the first line of treatment evidenced that those individuals treated with the best-predicted drug combination were significantly less likely to die than patients treated with other schemes. This was particularly important among patients treated with a triplet combination including bortezomib, an immunomodulatory drug (ImiD), and dexamethasone. Finally, the model showed a trend to retain its predictive value in patients with high-risk cytogenetics. In conclusion, we report a predictive model for MM survival based on the integration of clinical, biochemical, and gene expression data with machine learning tools.

Detection of new drivers of frequent B-cell lymphoid neoplasms using an integrated analysis of whole genomes.

B-cell lymphoproliferative disorders exhibit a diverse spectrum of diagnostic entities with heterogeneous behaviour. Multiple efforts have focused on the determination of the genomic drivers of B-cell lymphoma subtypes. In the meantime, the aggregation of diverse tumors in pan-cancer genomic studies has become a useful tool to detect new driver genes, while enabling the comparison of mutational patterns across tumors. Here we present an integrated analysis of 354 B-cell lymphoid disorders. 112 recurrently mutated genes were discovered, of which KMT2D, CREBBP, IGLL5 and BCL2 were the most frequent, and 31 genes were putative new drivers. Mutations in CREBBP, TNFRSF14 and KMT2D predominated in follicular lymphoma, whereas those in BTG2, HTA-A and PIM1 were more frequent in diffuse large B-cell lymphoma. Additionally, we discovered 31 significantly mutated protein networks, reinforcing the role of genes such as CREBBP, EEF1A1, STAT6, GNA13 and TP53, but also pointing towards a myriad of infrequent players in lymphomagenesis. Finally, we report aberrant expression of oncogenes and tumor suppressors associated with novel noncoding mutations (DTX1 and S1PR2), and new recurrent copy number aberrations affecting immune check-point regulators (CD83, PVR) and B-cell specific genes (TNFRSF13C). Our analysis expands the number of mutational drivers of B-cell lymphoid neoplasms, and identifies several differential somatic events between disease subtypes.

Detection of Rare Germline Variants in the Genomes of Patients with B-Cell Neoplasms.

There is growing evidence indicating the implication of germline variation in cancer predisposition and prognostication. Here, we describe an analysis of likely disruptive rare variants across the genomes of 726 patients with B-cell lymphoid neoplasms. We discovered a significant enrichment for two genes in rare dysfunctional variants, both of which participate in the regulation of oxidative stress pathways (CHMP6 and GSTA4). Additionally, we detected 1675 likely disrupting variants in genes associated with cancer, of which 44.75% were novel events and 7.88% were protein-truncating variants. Among these, the most frequently affected genes were ATM, BIRC6, CLTCL1A, and TSC2. Homozygous or germline double-hit variants were detected in 28 cases, and coexisting somatic events were observed in 17 patients, some of which affected key lymphoma drivers such as ATM, KMT2D, and MYC. Finally, we observed that variants in six different genes were independently associated with shorter survival in CLL. Our study results support an important role for rare germline variation in the pathogenesis and prognosis of B-cell lymphoid neoplasms.

Gene expression profiling identifies FLT3 mutation-like cases in wild-type FLT3 acute myeloid leukemia.

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.

Personally Tailored Survival Prediction of Patients With Follicular Lymphoma Using Machine Learning Transcriptome-Based Models.

Follicular Lymphoma (FL) has a 10-year mortality rate of 20%, and this is mostly related to lymphoma progression and transformation to higher grades. In the era of personalized medicine it has become increasingly important to provide patients with an optimal prediction about their expected outcomes. The objective of this work was to apply machine learning (ML) tools on gene expression data in order to create individualized predictions about survival in patients with FL. Using data from two different studies, we were able to create a model which achieved good prediction accuracies in both cohorts (c-indexes of 0.793 and 0.662 in the training and test sets). Integration of this model with m7-FLIPI and age rendered high prediction accuracies in the test set (cox c-index 0.79), and a simplified approach identified 4 groups with remarkably different outcomes in terms of survival. Importantly, one of the groups comprised 27.35% of patients and had a median survival of 4.64 years. In summary, we have created a gene expression-based individualized predictor of overall survival in FL that can improve the predictions of the m7-FLIPI score.

Improved personalized survival prediction of patients with diffuse large B-cell Lymphoma using gene expression profiling.

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.

The association of germline variants with chronic lymphocytic leukemia outcome suggests the implication of novel genes and pathways in clinical evolution.

BACKGROUND: Chronic Lymphocytic Leukemia (CLL) is the most frequent lymphoproliferative disorder in western countries and is characterized by a remarkable clinical heterogeneity. During the last decade, multiple genomic studies have identified a myriad of somatic events driving CLL proliferation and aggressivity. Nevertheless, and despite the mounting evidence of inherited risk for CLL development, the existence of germline variants associated with clinical outcomes has not been addressed in depth. METHODS: Exome sequencing data from control leukocytes of CLL patients involved in the International Cancer Genome Consortium (ICGC) was used for genotyping. Cox regression was used to detect variants associated with clinical outcomes. Gene and pathways level associations were also calculated. RESULTS: Single nucleotide polymorphisms in PPP4R2 and MAP3K4 were associated with earlier treatment need. A gene-level analysis evidenced a significant association of RIPK3 with both treatment need and survival. Furthermore, germline variability in pathways such as apoptosis, cell-cycle, pentose phosphate, GNα13 and Nitric oxide was associated with overall survival. CONCLUSION: Our results support the existence of inherited conditionants of CLL evolution and points towards genes and pathways that may results useful as biomarkers of disease outcome. More research is needed to validate these findings.

Time to Treatment Prediction in Chronic Lymphocytic Leukemia Based on New Transcriptional Patterns.

Chronic lymphocytic leukemia (CLL) is the most frequent lymphoproliferative syndrome in western countries. CLL evolution is frequently indolent, and treatment is mostly reserved for those patients with signs or symptoms of disease progression. In this work, we used RNA sequencing data from the International Cancer Genome Consortium CLL cohort to determine new gene expression patterns that correlate with clinical evolution.We determined that a 290-gene expression signature, in addition to immunoglobulin heavy chain variable region (IGHV) mutation status, stratifies patients into four groups with notably different time to first treatment. This finding was confirmed in an independent cohort. Similarly, we present a machine learning algorithm that predicts the need for treatment within the first 5 years following diagnosis using expression data from 2,198 genes. This predictor achieved 90% precision and 89% accuracy when classifying independent CLL cases. Our findings indicate that CLL progression risk largely correlates with particular transcriptomic patterns and paves the way for the identification of high-risk patients who might benefit from prompt therapy following diagnosis.