Molecular predictors of minimal residual disease in newly diagnosed multiple myeloma treated with quadruplet therapy.
TL;DR
A multiomic classification of NDMM linked to Isa-KRd response is defined, with a multimodal random forest model outperforming clinical and genomic models and providing comparable discrimination and improved calibration relative to the transcriptomic and epigenomic models.
Abstract
Quadruplet therapy, combining anti-CD38 antibodies, immunomodulatory drugs, proteasome inhibitors and dexamethasone, is now the standard induction treatment for newly diagnosed multiple myeloma (NDMM). However, treatment response is heterogeneous and molecular predictors are needed to guide early therapeutic decisions. We performed integrated genomic, transcriptomic and epigenomic profiling of 423 tumor samples from patients with NDMM from the MIDAS cohort to characterize molecular heterogeneity and its association with measurable residual disease (MRD) after quadruplet induction with isatuximab, carfilzomib, lenalidomide and dexamethasone (Isa-KRd). Postinduction MRD-negativity rates varied significantly across gene expression subtypes, with HP1/HP3 and MS groups showing superior responses compared to the MF and CD2 subtypes. B cell differentiation was associated with MRD positivity, while proliferation and MYC programs were associated with MRD negativity. These MRD-associated transcriptomic features were independently validated in the CASSIOPEIA cohort. Whole-genome sequencing revealed both known (APOBEC mutagenesis, chromothripsis) and new genomic instability phenotypes validated in the MMRF cohort. New driver alterations were identified with independent validation across three external cohorts with multiple myeloma (MMRF, MANHATTAN and CASSIOPEIA; 1,451 cases). Genomic instability features and driver genes such as CCND1 and MYC correlated with differential MRD outcomes. DNA methylation profiling identified nine MRD-linked subgroups distinguished by the extent of hypomethylation in quiescent chromatin and cis/trans signatures of driver translocations. A multimodal random forest model integrating genomic, transcriptomic and epigenomic features predicted MRD status after induction, with an area under the curve of 0.77, outperforming clinical and genomic models, while providing comparable discrimination and improved calibration relative to the transcriptomic and epigenomic models. These findings define a multiomic classification of NDMM linked to Isa-KRd response. ClinicalTrials.gov identifier: NCT04934475 .