The investigators analyzed 423 tumor samples from patients with newly diagnosed multiple myeloma enrolled in the MIDAS cohort, all treated with quadruplet induction containing isatuximab, carfilzomib, lenalidomide and dexamethasone (Isa-KRd). Response to induction showed clear heterogeneity. Postinduction minimal residual disease (MRD) negativity was not uniform across molecular groups; the HP1/HP3 and MS gene-expression subtypes exhibited superior MRD-negativity rates compared with the MF and CD2 subtypes, the authors reported. The cohort therefore stratified by transcriptomic subtype into groups with distinct early depth-of-response profiles.
The study team observed that transcriptional programs associated with B cell differentiation correlated with MRD positivity. By contrast, signatures linked to proliferation and MYC activity associated with MRD negativity after induction. These contrasting transcriptomic features were tested against an external dataset: the CASSIOPEIA cohort. Data indicated that the MRD-associated transcriptomic signals identified in MIDAS were reproduced in that independent trial, supporting the robustness of expression-based predictors of early response to quadruplet therapy.
Whole-genome sequencing in the MIDAS samples identified established mutational processes known in multiple myeloma, including APOBEC-related mutagenesis and chromothripsis. The investigators also described additional genomic instability phenotypes that had not been as widely characterized; those new instability patterns were validated in the MMRF cohort. Across analyses, genomic instability features and specific driver gene alterations correlated with differential MRD outcomes, indicating that structural and mutational processes in the tumor genome influence early treatment sensitivity to Isa-KRd.
Genome-wide DNA methylation profiling partitioned tumors into nine MRD-linked epigenetic subgroups. These groups were distinguished primarily by the extent of hypomethylation in quiescent chromatin compartments and by cis and trans signatures associated with driver translocations. The methylation patterns therefore provided an additional, orthogonal molecular axis associated with MRD status after induction. The authors framed the methylome as a contributor to transcriptional plasticity that maps onto measurable treatment response.
A multimodal random forest model was constructed to predict MRD after induction by integrating genomic, transcriptomic and epigenomic features. The integrated model yielded an area under the curve of 0.77. The investigators compared this model to alternatives: it outperformed clinical-only and genomic-only models, offered comparable discrimination to transcriptomic- and epigenomic-only models, and delivered improved calibration relative to models that used single molecular layers. The report provides the model AUC as a quantitative benchmark of multimodal prediction performance in this treatment context.
Beyond internal findings, the study team validated newly identified driver alterations across three external multiple-myeloma datasets: MMRF, MANHATTAN and CASSIOPEIA, together encompassing 1,451 cases. Known drivers and the newly highlighted alterations — with specific mention of genes such as CCND1 and MYC — showed associations with MRD outcomes. Validation across multiple, independent cohorts strengthened the evidence that certain genomic events and instability phenotypes consistently influence early depth of response to quadruplet regimens.
The authors emphasized a previously described shortfall in the field: large, prospectively collected trials that integrate genome, transcriptome and methylome data were lacking prior to this effort. This study addresses that gap by linking multiomic classifiers to response after modern quadruplet induction. The SOURCE did not report granular treatment-dosing schedules, long-term survival outcomes beyond postinduction MRD, or regulatory next steps. Open questions remain about how these multiomic classifiers should be applied in clinical decision-making, whether they predict long-term progression-free or overall survival, and how the predictive model will perform in routine clinical settings. These issues outline concrete areas for follow-up research and prospective clinical validation. ClinicalTrials.gov identifier: NCT04934475.