Abstract 3773: DeMixMir: deconvolution of microRNA sequencing data from heterogeneous tumor samples

Abstract

Abstract Measuring tumor heterogeneity is a key issue in modern clinical oncology, with relevance to understanding cancer progression, resistance to therapy, and recurrence. Studies of the transcriptomic landscape in cancer have rapidly advanced in scale, enabled by novel methods and advancing technologies. One such novel method is TmS, a computational estimate of tumor-specific total messenger RNA content from heterogeneous tumor samples, which is calculated by combining genomic and transcriptomic deconvolutions. We have previously shown TmS as a promising pan-cancer biomarker for patient prognosis. Given the fruitful study of tumor cell total messenger RNA content, a logical extension is to investigate the utility of estimating tumor cell content of other RNA species. MicroRNAs (miRNAs) are small noncoding RNAs that regulate mRNA expression, and their dysregulation is a hallmark feature observed across cancers. A computational method to derive tumor cell total miRNA content from bulk sequencing data is especially needed since miRNAs cannot yet be reliably profiled at single cell resolution. Here we develop and benchmark DeMixMir, a new expansion of our reference-free transcriptomic deconvolutional model DeMixT, in order to recover tumor-specific miRNA proportions from mixed samples. For benchmarking, we generated an artificially mixed dataset of small RNA sequencing consisting of a total of 30 samples that were made by mixing HS-5 fibroblast cells with either wildtype or Dicer1 knockout HCT116 colorectal cancer cells. The tumor and fibroblast cell lines were mixed at 5 different proportions, to simulate a broad spectrum of tumor/nontumor cell mixing scenarios, with three independent replicates generated at each mixing ratio. DeMixMir demonstrated high accuracy in estimating the tumor-specific miRNA proportions in both the wildtype and the Dicer1 knockout mixtures. We further calculated tumor cell total miRNA content using DeMixMir output. Our proof-of-concept study suggests estimating tumor cell total miRNA content across tumor tissues at scale is feasible and likely valuable for advancing understanding of the complex roles miRNAs play in the cancer ecosystem. Citation Format: Matthew Montierth, Kinga Nemeth, Xinghua Tao, George Calin, Wenyi Wang. DeMixMir: deconvolution of microRNA sequencing data from heterogeneous tumor samples. [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 3773.