Abstract
Abstract Patient derived xenografts (PDXs) without in vitro manipulation are believed to mirror original patients’ histopathologic and genetic profiles, thus to be predictive surrogate models for patients, with superiority over conventional cell lines. We have built the largest and comprehensive PDX library of >1,100 models with genetic profiles of major cancer types, including some major cancer types: 200 NSLCL1, 200 CRC2, 200 gastric3, 100 HCC4, 100 pancreatic, 30 ovarian, 10 brain tumors, and many other cancer types. We set out to compare these 6 types of our PDXs with the corresponding TCGA5 patient tumor samples and CCLE6 cancer cell lines on their genomic expression by calculating pairwise Spearman rank correlation coefficient ρ, in order to further explore/confirm the similarity and difference among the three collections. PDX were profiled by both RNAseq and microarray (Affymetrix Human Genome U219 Array); CCLE were from microarray (Affymetrix Human Genome U133 Plus 2.0 Array); and TCGA were from RNAseq. For convenience, these 4 gene expression datasets are called “PDX, PDXchip, CCLE and TCGA”. Only genes common to all 4 datasets were used to compute ρ. The preliminary data seems to lead to two observations. First, for all 6 types, there is always at least one different tumor CCLE-TCGA pair that has higher ρ than the same tumor CCLE-TCGA pair. For example, the TCGA colorectal, lung, and ovarian cancers have higher ρ with CCLE pancreatic cancer than the TCGA pancreatic cancer does. In contrast, for both PDX and PDXchip, only lung and pancreatic cancers show such behavior, and notable, both to TCGA colorectal cancer. Seocnd, for same tumor type comparisons, ρ is 0.73-0.83 (average 0.772) between PDX and TCGA, 0.67-0.76 (average 0.698) between PDXchip and TCGA, 0.67-0.7 (average 0.682) between CCLE and TCGA. Third, about 20% of CCLE lung cancers have much lower ρ with TCGA, PDX, and PDXchip lung cancers than the other CCLE lung cancers. In summary, these observations show that our PDX models are close in genomic expression profile to TCGA patient tumors per tumor types specifically, and more so than CCLE cancer cell lines, which is also less specific.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have