Background: Cancers are heterogeneous diseases, requiring individualized treatment approaches and relying on predictive biomarkers. Gastric carcinoma (GC), a common cancer with high fatality, particularly in East Asia, urgently needs effective treatment options, particularly targeted therapy. Patient-derived xenografts (PDXs) are widely accepted as predictive models for translational research, including identifying predictive biomarkers to guide GC-treatment (Zhang et al., 2013). The comprehensive annotation of PDXs pathologically, genomically and proteomically, at baseline and/or upon treatment (pharmacodynamics, or PD), are essential for biomarker discovery and validation. We previously described establishment of a panel of GC-PDX with transcriptome annotation and utilized these models for biomarker discovery predictive of the cetuximab response (Zhang et al., 2013). The present study aims to comparing the genomic and proteomic profiles of the GC-PDX panel and their, usage in translational research. Method. 48 GC-PDX tumor are subjected to transcriptome (RNASeq) sequencing and whole proteomic analysis using high-throughput label-free MS-based technology. We calculate spearman’s correlation coefficient (cc) between relative protein abundance metric (ifot) and mRNA expression metric (FPKM) per proteomic datasets and RNASeq datasets, respectively, to assess their consistent/differential features. We also tested an integrated -omics approach by DIABLO algorithm (Singh et al., 2017) for predictive biomarker discovery using mouse clinical trial data. Results. The correlation between mRNA and protein abundance was moderate (Spearman’s cc~0.55) consistent across all tested PDXs. Genes with high mRNA expression levels have weaker correlations with their corresponding protein abundance. This could be attributed to stronger post-transcriptional control on protein expression for these highly transcribed genes. Gene ontology (GO) over representation analysis were conducted for two subsets of genes: 1) ribosome and cadherin binding related proteins that are highly expressed at both mRNA and protein level, while 2) histones and other DNA binding genes have high protein abundance but low mRNA expression. We also conducted predictive biomarker analysis using the cetuximab (an EGFR inhibitor) GC-PDX trial data, where we previously demonstrated high correlation between EGFR mRNA expression and sensitivity to cetuximab. Although EGFR protein abundance is positively correlated with cetuximab efficacy, the correlation seems weaker than for their mRNA counterpart in this case. We also used the DIABLO algorithm for the integration of both proteomics and transcriptome data, which aims to identify a multi-omics signature that is maximally correlated across data types. A signature including 10-mRNA expression/5-protein abundance has been identified, with high inter data type correlation (0.94)/good discrimination power. Conclusion. The variability of protein abundance can be partially explained by mRNA level, however, post-transcriptional regulation could play important role on protein expression especially for highly transcribed genes. Proteomics data may provide complimentary information to transcriptome data for predictive biomarker analysis.

Citation Format: Binchen Mao, Sheng Guo, Bonnie Xiaobo Chen, Davy Xuesong Ouyang, Henry Li. Comparative analysis of proteomic/transcriptomic profiles of a panel of GC-PDXs [abstract]. In: Proceedings of the AACR-NCI-EORTC International Conference on Molecular Targets and Cancer Therapeutics; 2019 Oct 26-30; Boston, MA. Philadelphia (PA): AACR; Mol Cancer Ther 2019;18(12 Suppl):Abstract nr A025. doi:10.1158/1535-7163.TARG-19-A025