Background: Clinical management and research of multiple myeloma (MM), a cancer of plasma cells, is limited by tumor heterogeneity. A standard approach to deconstruct tumor heterogeneity is to use hierarchical clustering techniques to determine mutually exclusive categorical subtypes. However, categorical subtypes may fail to capture potential important variations that cross tumor subtypes. An alternate approach is to determine orthogonal, quantitative tumor dimensions where each dimension is an independent tumor characteristic. We hypothesize that using a quantitative framework for tumor heterogeneity in MM will uncover biologically relevant components to tumors and may reflect specific molecular liabilities and therapeutic vulnerabilities. Further, these quantitative characteristics may be more homogeneous genetically and useful for germline gene mapping.

Objective: Identify orthogonal, quantitative dimensions in MM tumors using gene expression.

Data: RNA sequencing on treatment-naïve, CD138 sorted tumor cells from 768 individuals. Publicly available from the MM Research Foundation’s Clinical Outcomes on MM Genetic Profiles Assessment (CoMMpass) Interim Analysis 12a. SALMON transcripts per million adjusted expression estimates on 16,870 protein coding genes.

Analyses: Multistage singular value decomposition (SVD) to 1) select representative genes, and 2) characterize the orthogonal, quantitative tumor dimensions. Stage 1: Genes that contribute most to the initial SVD will be selected as representative genes. Stage 2: SVD on selected genes to identify quantitative gene expression tumor dimensions. Each dimension is a linear combination of the representative genes. Future work will associate the quantitative dimensions with demographic, clinical, and genetic (germline and somatic) characteristics, in addition to response to treatment using penalized linear regression modeling.

Conclusions: We present a new approach for the characterization of MM tumors using a more sophisticated quantitative framework that will facilitate more flexibility for subsequent statistical modeling. Improved measures for tumors have the potential to provide increased power for identification of association between tumor characteristics and genetic (germline or somatic) characteristics with ultimate potential for genetic counseling, insights into mechanism, risk stratification, response to treatment, and new candidates for precision therapeutics.

Citation Format: Rosalie G. Waller, Michael J. Madsen, John Gardner, Douglas Sborov, Nicola J. Camp. Characterization of quantitative gene-expression dimensions in myeloma tumors [abstract]. In: Proceedings of the AACR Special Conference on Modernizing Population Sciences in the Digital Age; 2019 Feb 19-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Epidemiol Biomarkers Prev 2020;29(9 Suppl):Abstract nr A40.