MOSAIC: An Artificial Intelligence-Based Framework for Multimodal Analysis, Classification, and Personalized Prognostic Assessment in Rare Cancers.

PURPOSE: Rare cancers constitute over 20% of human neoplasms, often affecting patients with unmet medical needs. The development of effective classification and prognostication systems is crucial to improve the decision-making process and drive innovative treatment strategies. We have created and implemented MOSAIC, an artificial intelligence (AI)-based framework designed for multimodal analysis, classification, and personalized prognostic assessment in rare cancers. Clinical validation was performed on myelodysplastic syndrome (MDS), a rare hematologic cancer with clinical and genomic heterogeneities. METHODS: We analyzed 4,427 patients with MDS divided into training and validation cohorts. Deep learning methods were applied to integrate and impute clinical/genomic features. Clustering was performed by combining Uniform Manifold Approximation and Projection for Dimension Reduction + Hierarchical Density-Based Spatial Clustering of Applications with Noise (UMAP + HDBSCAN) methods, compared with the conventional Hierarchical Dirichlet Process (HDP). Linear and AI-based nonlinear approaches were compared for survival prediction. Explainable AI (Shapley Additive Explanations approach [SHAP]) and federated learning were used to improve the interpretation and the performance of the clinical models, integrating them into distributed infrastructure. RESULTS: UMAP + HDBSCAN clustering obtained a more granular patient stratification, achieving a higher average silhouette coefficient (0.16) with respect to HDP (0.01) and higher balanced accuracy in cluster classification by Random Forest (92.7% ± 1.3% and 85.8% ± 0.8%). AI methods for survival prediction outperform conventional statistical techniques and the reference prognostic tool for MDS. Nonlinear Gradient Boosting Survival stands in the internal (Concordance-Index [C-Index], 0.77; SD, 0.01) and external validation (C-Index, 0.74; SD, 0.02). SHAP analysis revealed that similar features drove patients' subgroups and outcomes in both training and validation cohorts. Federated implementation improved the accuracy of developed models. CONCLUSION: MOSAIC provides an explainable and robust framework to optimize classification and prognostic assessment of rare cancers. AI-based approaches demonstrated superior accuracy in capturing genomic similarities and providing individual prognostic information compared with conventional statistical methods. Its federated implementation ensures broad clinical application, guaranteeing high performance and data protection.

A geno-clinical decision model for the diagnosis of myelodysplastic syndromes.

The differential diagnosis of myeloid malignancies is challenging and subject to interobserver variability. We used clinical and next-generation sequencing (NGS) data to develop a machine learning model for the diagnosis of myeloid malignancies independent of bone marrow biopsy data based on a 3-institution, international cohort of patients. The model achieves high performance, with model interpretations indicating that it relies on factors similar to those used by clinicians. In addition, we describe associations between NGS findings and clinically important phenotypes and introduce the use of machine learning algorithms to elucidate clinicogenomic relationships.

Classification and Personalized Prognostic Assessment on the Basis of Clinical and Genomic Features in Myelodysplastic Syndromes.

PURPOSE: Recurrently mutated genes and chromosomal abnormalities have been identified in myelodysplastic syndromes (MDS). We aim to integrate these genomic features into disease classification and prognostication. METHODS: We retrospectively enrolled 2,043 patients. Using Bayesian networks and Dirichlet processes, we combined mutations in 47 genes with cytogenetic abnormalities to identify genetic associations and subgroups. Random-effects Cox proportional hazards multistate modeling was used for developing prognostic models. An independent validation on 318 cases was performed. RESULTS: We identify eight MDS groups (clusters) according to specific genomic features. In five groups, dominant genomic features include splicing gene mutations (SF3B1, SRSF2, and U2AF1) that occur early in disease history, determine specific phenotypes, and drive disease evolution. These groups display different prognosis (groups with SF3B1 mutations being associated with better survival). Specific co-mutation patterns account for clinical heterogeneity within SF3B1- and SRSF2-related MDS. MDS with complex karyotype and/or TP53 gene abnormalities and MDS with acute leukemia-like mutations show poorest prognosis. MDS with 5q deletion are clustered into two distinct groups according to the number of mutated genes and/or presence of TP53 mutations. By integrating 63 clinical and genomic variables, we define a novel prognostic model that generates personally tailored predictions of survival. The predicted and observed outcomes correlate well in internal cross-validation and in an independent external cohort. This model substantially improves predictive accuracy of currently available prognostic tools. We have created a Web portal that allows outcome predictions to be generated for user-defined constellations of genomic and clinical features. CONCLUSION: Genomic landscape in MDS reveals distinct subgroups associated with specific clinical features and discrete patterns of evolution, providing a proof of concept for next-generation disease classification and prognosis.

Integrating clinical features and genetic lesions in the risk assessment of patients with chronic myelomonocytic leukemia.

Chronic myelomonocytic leukemia (CMML) is a myelodysplastic/myeloproliferative neoplasm with variable clinical course. To predict the clinical outcome, we previously developed a CMML-specific prognostic scoring system (CPSS) based on clinical parameters and cytogenetics. In this work, we tested the hypothesis that accounting for gene mutations would further improve risk stratification of CMML patients. We therefore sequenced 38 genes to explore the role of somatic mutations in disease phenotype and clinical outcome. Overall, 199 of 214 (93%) CMML patients carried at least 1 somatic mutation. Stepwise linear regression models showed that these mutations accounted for 15% to 24% of variability of clinical phenotype. Based on multivariable Cox regression analyses, cytogenetic abnormalities and mutations in RUNX1, NRAS, SETBP1, and ASXL1 were independently associated with overall survival (OS). Using these parameters, we defined a genetic score that identified 4 categories with significantly different OS and cumulative incidence of leukemic evolution. In multivariable analyses, genetic score, red blood cell transfusion dependency, white blood cell count, and marrow blasts retained independent prognostic value. These parameters were included into a clinical/molecular CPSS (CPSS-Mol) model that identified 4 risk groups with markedly different median OS (from >144 to 18 months, hazard ratio [HR] = 2.69) and cumulative incidence of leukemic evolution (from 0% to 48% at 4 years, HR = 3.84) (P < .001). The CPSS-Mol fully retained its ability to risk stratify in an independent validation cohort of 260 CMML patients. In conclusion, integrating conventional parameters and gene mutations significantly improves risk stratification of CMML patients, providing a robust basis for clinical decision-making and a reliable tool for clinical trials.

Mutational profiling in patients with MDS: ready for every-day use in the clinic?

Multiple recurrent somatic mutations were identified in the majority of patients with myelodysplastic syndromes (MDS), but investigating the broad spectrum of molecular markers in MDS exceeds many laboratories' capacity when traditional molecular techniques are used. High-throughput second generation sequencing (=next-generation sequencing, NGS) has proven to be applicable for comprehensive biomarker mutation analyses allowing to increase diagnostic sensitivity and accuracy and to improve risk stratification and prognostication in addition to cytomorphology and cytogenetic analysis in patients with MDS. Amplicon deep-sequencing enables comprehensive biomarker analysis in a multitude of patients per investigation in an acceptable turn-around time and at affordable costs. Comprehensive myeloid marker panels were successfully introduced into diagnostic practice. Therefore, molecular mutation analysis is ready for use in all patients with suspected MDS, may contribute to risk stratification in possible candidates for allogeneic stem cell transplantation, and should become an integral part of clinical research studies in MDS patients.

Relapse assessment following allogeneic SCT in patients with MDS and AML.

Options to pre-emptively treat impending relapse of myelodysplastic syndromes (MDS) and acute myeloid leukaemia (AML) after allogeneic haematopoietic stem cell transplantation (allo-SCT) continuously increase. In recent years, the spectrum of diagnostic methods and parameters to perform post-transplant monitoring in patients with AML and MDS has grown. Cytomorphology, histomorphology, and chimaerism analysis are the mainstay in any panel of post-transplant monitoring. This may be individually combined with multiparameter flow cytometry (MFC) for the detection of residual cells with a leukaemia phenotype and quantitative real-time polymerase chain reaction (RQ-PCR) to assess gene expression, e.g., of WT1 or the residual mutation load (e.g., in case of an NPM1 mutation). Data evaluating the aforementioned methods alone or in combination are discussed in this review with particular emphasis on data pointing towards their suitability to steer pre-emptive post-transplant interventions such as immunotherapy, chemotherapy or therapy with demethylating agents.

Serial assessment of suspected myelodysplastic syndromes: significance of flow cytometric findings validated by cytomorphology, cytogenetics, and molecular genetics.

The significance of flow cytometry indicating myelodysplasia without proof of myelodysplasia by cytomorphology remains to be clarified. We evaluated follow-up analyses in 142 patients analyzed in parallel by flow cytometry, cytomorphology and cytogenetics for suspected myelodysplasia without proof of myelodysplasia by cytomorphology. At initial assessment, flow cytometry indicated myelodysplasia in 64 of 142 (45.1%) patients. In 9 of 142 (6.3%) patients, cytogenetics revealed aberrant karyotypes at first evaluation that were found in 5 of 64 (7.8%) patients rated with myelodysplasia by flow cytometry. The remaining 133 patients without proof of myelodysplasia by cytomorphology and with normal karyotype underwent follow-up analyses that confirmed myelodysplasia by cytomorphology, cytogenetics or molecular genetics in 47 (35.3%) after a median interval of nine months (range 1-53 months). As far as initial flow cytometry results are concerned, this applied to 30 of 59 (50.1%) with myelodysplasia, 10 of 42 (23.8%) with "possible myelodysplasia" (minor antigen aberrancies only) and 7 of 32 (21.9%) without myelodysplasia (P=0.004). Notably, in these latter 7 patients, flow cytometry results changed at follow up to "possible myelodysplasia" (n=4) and "myelodysplasia" (n=2). These data argue in favor of including flow cytometry along with cytomorphology, cytogenetics and molecular genetics to diagnose myelodysplasia, and suggest a closer monitoring of patients with myelodysplasia-typical aberrant antigen expression found by flow cytometry.

Evaluation of flow cytometric assessment of myeloid nuclear differentiation antigen expression as a diagnostic marker for myelodysplastic syndromes in a series of 269 patients.

BACKGROUND: Myeloid nuclear differentiation antigen (MNDA) is expressed in myelomonocytic cells with highest levels in mature granulocytes and monocytes. It is suggested to be expressed more weakly in patients with myelodysplastic syndromes (MDS). The analysis of MNDA therefore may improve diagnostic capabilities of multiparameter flow cytometry (MFC) in MDS. METHODS: We used MFC for detection of MNDA expression in 269 patients with suspected or known MDS, acute myeloid leukemia (AML) or chronic myelomonocytic leukemia (CMML), cytopenia of unknown cause or without malignancy (negative controls). Results were compared with the diagnoses revealed by cytomorphology (CM) and cytogenetics (CG). RESULTS: Percentages of granulocytes and monocytes with diminished MNDA expression (dimG and dimM) were higher in patients with MDS (mean ± SD, 20% ± 20%, P < 0.001 and 31% ± 24%, P < 0.001) and AML (27% ± 27%, P = 0.007 and 45% ± 31%, P = 0.001) diagnosed by CM, vs. patients without MDS (8% ± 10% and 16% ± 11%), respectively. Significant differences were also found for mean fluorescence intensity (MFI) of MNDA in monocytes which was lower in MDS (mean ± SD, 71 ± 36, P = 0.004) and AML (55 ± 39, P < 0.001) vs. no MDS samples (85 ± 28), respectively. Within patients with MDS, cases with cytogenetic aberrations showed a trend to higher %dimG (24% ± 18%, P = 0.083) compared with those without (16% ± 21%). Cut-off values for %dimG (12%) and %dimM (22%) as well as for MFI in monocytes (72) were defined capable of discriminating between MDS and non-MDS. CONCLUSION: MNDA expression in bone marrow cells can be assessed reliably by MFC and may facilitate evaluation of dyspoiesis when added to a standard MDS MFC panel.

Risk assessment by monitoring expression levels of partial tandem duplications in the MLL gene in acute myeloid leukemia during therapy.

BACKGROUND AND OBJECTIVES: A partial tandem duplication within the MLL-gene (MLL-PTD) can be found in 8% of all patients with karyotypically normal acute myeloid leukemia (AML), a group in which polymerase chain reaction-(PCR) based minimal residual disease analysis has not, so far, been possible. DESIGN AND METHODS: A sensitive real-time PCR assay to quantify MLL-PTD transcripts was established and expression ratios assessed in diagnostic and follow-up samples. The prognostic significance of MLL-PTD expression levels was evaluated in 145 MLL-PTD positive patients at diagnosis and in 44 patients during and after treatment. RESULTS: Paired samples from 16 patients evaluated at diagnosis and relapse for the presence of the MLL-PTD were analyzed in parallel and all samples were positive at both time points. Overall, 173 samples from 44 patients were analyzed during follow-up (median sample number: 4/patient (range 2-17)). Nineteen patients were evaluable for MRD within the first 2 months, 15 patients within 4 months, and 19 patients within 6 months after the start of therapy. A >or= 2 log reduction of MLL-PTD expression in comparison to < 2 log reduction within 2, 4, and 6 months after start of therapy was found to be significantly associated with longer overall survival (p=0.029, p=0.007, and p=0.022, respectively). A molecular relapse was detected in 2 cases, in each case preceeding clinical manifestation by 35 days. INTERPRETATION AND CONCLUSIONS: These data suggest that MLL-PTD is a stable marker and can be used as a prognostically important marker of MRD in patients with karyotypically normal AML.

Four-fold staining including CD45 gating improves the sensitivity of multiparameter flow cytometric assessment of minimal residual disease in patients with acute myeloid leukemia.

Multiparameter flow cytometry (MFC) is capable of quantifying minimal residual disease (MRD) in acute myeloid leukemia (AML). Its broad application, however, is limited by a lack of sensitivity in about 20% of patients. CD45 gating may improve sensitivity. A broad panel of four-fold combinations of monoclonal antibodies including CD45 in each was used to define leukemia-associated aberrant immunophenotypes (LAIP), to define their sensitivities in normal bone marrow samples, and to compare results to data obtained without CD45 gating using triple staining. In 45 patients, a LAIP was defined, 11 normal bone marrow samples were analyzed as controls. The median percentage of LAIP-positive AML cells with and without CD45 gating was 21.95% (range, 3.31-82.52%) and 20.52% (range, 3.22-81.94%). The median percentage of LAIP-positive normal bone marrow cells ranged from 0.01 to 0.42% (median, 0.02%) and 0.02 to 0.58% (median, 0.15%) with and without CD45 gating. The difference of LAIP-positive cells between AML and normal bone marrow samples amounted to a median of 3.08 log (range, 1.22-4.01) and 2.28 log (range, 1.12-3.34) with and without CD45 gating. CD45 gating improves the sensitivity of MFC-based MRD monitoring in AML by 1 log.

Determination of relapse risk based on assessment of minimal residual disease during complete remission by multiparameter flow cytometry in unselected patients with acute myeloid leukemia.

Quantification of minimal residual disease (MRD) reveals significant prognostic information in patients treated for acute myeloid leukemia (AML). The application of multiparameter flow cytometry (MFC) for MRD assessment has resulted in significant prognostic information in selected cases in previous analyses. We analyzed MRD in unselected patients with AML in complete remission (CR) after induction (n = 58) and consolidation (n = 62) therapies. By using a comprehensive panel of monoclonal antibodies we identified at least one leukemia-associated aberrant immunophenotype (LAIP) in each patient. The degree of reduction between diagnosis and CR in LAIP-positive cells (log difference [LD]) as a continuous variable was significantly related to relapse-free survival (RFS) both after induction (P = .0001) and consolidation (P = .000 08) therapies, respectively. The LD determined after consolidation therapy was the only parameter related to overall survival (OS) (P = .005). Separation of patients based on the 75th percentile of LD after consolidation therapy resulted in groups with highly different RFS (83.3% versus 25.7%, P = .0034) and OS (87.5% versus 51.4%, P = .0507) at 2 years. Multivariate analysis identified LD as an independent prognostic factor for RFS at both checkpoints. MFC-based quantification of MRD reveals important prognostic information in unselected patients with AML in addition to cytogenetics and should be further evaluated and used in clinical trials.