(PA-260) Independent and Complementary Value of RNA Expression Signatures in High-Risk Multiple Myeloma

Gene expression signatures are valuable for capturing the biological heterogeneity of multiple myeloma (MM), revealing disease mechanisms, progression, and treatment vulnerabilities. However, frequent somatic changes in MM, such as copy number alterations and translocations, impact the transcriptome. To evaluate the added value of RNA-based information, we developed an ensemble of RNA signatures and compared them to DNA-based models like the IMS/IMWG Consensus Genomic Staging for high-risk MM.


Methods:

We analyzed RNA sequencing data from 1,620 newly diagnosed MM patients and added 690 with array-based profiles, totaling 2,310 patients across five independent datasets. Seven established gene expression risk models were harmonized for RNAseq platforms, and their combined risk scores were stratified into four categories based on data type.


Results:

We compared both continuous and stratified gene expression (GEX) risk scores—GEX-negative, GEX high-risk (≥4 positive signatures), GPI-positive, and Others (1–3 signatures, GPI-negative)—against a DNA-based risk model. Across all datasets, GEX high-risk and GPI-positive patients showed significantly worse progression-free survival (PFS; HR 3.6 and 2.6) and overall survival (OS; HR 4.45 and 3.97), with p-values < 1.0e-6.

We then compared the RNA ensemble to the IMS/IMWG risk model. In our multivariate analysis, we discovered that both the DNA-based (HR=1.4, p< 0.001) and RNA-based (HR=3.6, p< 0.001) risk models serve as independent risk factors. The RNA-based models identified an additional 7% of patients within the IMS/IMWG risk stratification who had significantly poorer outcomes, while there were no corresponding DNA markers evident.

To explore mechanisms driving risk in GEX high-risk patients without GPI, we compared their DNA profiles to other groups. While GEX high-risk and GPI groups had similar frequencies of key DNA risk features, patients negative for all signatures—who had better outcomes—showed significantly fewer classical DNA risk markers, including TP53 mutations, 1q gains, and CDKN2C deletions (FDR < 0.01).

We further analyzed transcriptomic features unique to the GEX high-risk group lacking GPI and found significant differences from the other groups. A total of 482 genes were dysregulated (adj p < 0.01), with strong enrichment in cell adhesion pathways (FDR < 0.01). This enrichment was validated across all datasets using gene set variation analysis, confirming cell adhesion as a hallmark of this subgroup (Kruskal-Wallis p < 1e-5).


Conclusions:

Our analysis of a large patient cohort across multiple datasets demonstrates the potential to establish a standardized workflow for combining gene expression signatures. It also highlights that some patients cannot be accurately identified using current DNA risk markers but still have poor prognoses. Therefore, RNA expression retains independent information that should be considered in future studies.