Abstract 3168: DE-meta: Revealing tumor gene expression by meta-analysis of RNASeq and proteomics datasets

Abstract

Abstract Detecting differentially expressed genes under different biological conditions is crucial to characterize mechanisms of cancer development and identifying determinants of patient outcome. High throughput technologies such as RNA sequencing and mass spectrometry-based proteomics have been widely used to identify differentially expressed genes (DEG), on a transcript and on a peptide basis, respectively. However, both transcription and translation of genes provide information about gene expression. Thus, leveraging both RNA and protein expression data could potentially produce more accurate results. Various statistical tools have been developed to tackle the differential expression problem for a single platform, such as edgeR, DESeq2, etc. However, a tool that integrates both transcriptome and proteomics data for differential expression analysis has not yet been developed. Meta-analysis can potentially increase statistical power and generate a more consensus conclusion by combining multiple datasets. Here we present DE-Meta, a new tool developed using R, which performs combined meta-analysis of RNA-seq and MS-based proteomics on matched tumor specimens. To characterize the performance of our method using real world data, we utilized a published lung cancer dataset and set a truth standard of DEG between known two expression subtypes: terminal respiratory unit (TRU) and proximal proliferative (PP). Briefly, we restricted to 7,458 common genes detected by RNA-seq and MS-based proteomics. We then, randomly selected tumors as the truth sample set, n=20 TRU and n=20 PP tumors. With this truth sample set, DEGs were calculated separately by RNA or by protein, using DESeq2 and regression models. The union of these DEGs defined the truth DEG set. Using independent tumors (n=10 TRU and n=10 PP), we applied DE-meta to detect DEGs by RNA, by protein or joint RNA and protein data. Through receiver operating characteristic curve analysis, we show that the DE-meta joint model performed better than the single platform models (joint AUC=0.651, RNA AUC=0.610, protein AUC=0.623). Focusing genes relevant to tumor biology, we found that the DE-meta joint model called an 169 cancer-relevant DEGs, such as MSH2, CD4, WDR3, PIK3R1, and RPS6KA3. which were not detected by the RNA or protein models, at the same false discovery rate threshold of 0.05 (Benjamini-Hochberg procedure). In summary, our new DE-meta tool, by joint analysis of RNA-seq and MS-based proteomics datasets, yields more accurate and biologically-relevant results. The views expressed in this abstract are solely of the authors and do not reflect the official policy of the Departments of Army/Navy/Air Force, Department of Defense, USUHS, HJF, or U.S. Government. Citation Format: Xijun Zhang, Clifton L. Dalgard, Matthew D. Wilkerson. DE-meta: Revealing tumor gene expression by meta-analysis of RNASeq and proteomics datasets [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 3168.