Evaluating the transcriptional fidelity of cancer models

Abstract

Background Cancer researchers use cell lines, patient-derived xenografts, engineered mice, and tumoroids as models to investigate tumor biology and to identify therapies. The generalizability and power of a model derive from the fidelity with which it represents the tumor type under investigation; however, the extent to which this is true is often unclear. The preponderance of models and the ability to readily generate new ones has created a demand for tools that can measure the extent and ways in which cancer models resemble or diverge from native tumors. Methods We developed a machine learning-based computational tool, CancerCellNet, that measures the similarity of cancer models to 22 naturally occurring tumor types and 36 subtypes, in a platform and species agnostic manner. We applied this tool to 657 cancer cell lines, 415 patient-derived xenografts, 26 distinct genetically engineered mouse models, and 131 tumoroids. We validated CancerCellNet by application to independent data, and we tested several predictions with immunofluorescence. Results We have documented the cancer models with the greatest transcriptional fidelity to natural tumors, we have identified cancers underserved by adequate models, and we have found models with annotations that do not match their classification. By comparing models across modalities, we report that, on average, genetically engineered mice and tumoroids have higher transcriptional fidelity than patient-derived xenografts and cell lines in four out of five tumor types. However, several patient-derived xenografts and tumoroids have classification scores that are on par with native tumors, highlighting both their potential as faithful model classes and their heterogeneity. Conclusions CancerCellNet enables the rapid assessment of transcriptional fidelity of tumor models. We have made CancerCellNet available as a freely downloadable R package (https://github.com/pcahan1/cancerCellNet) and as a web application (http://www.cahanlab.org/resources/cancerCellNet_web) that can be applied to new cancer models that allows for direct comparison to the cancer models evaluated here.

CNpare: matching DNA copy number profiles

“…CNpare is the first stand-alone tool to facilitate comparison of cell line models based on high-resolution, genome-wide copy number. This complements existing approaches based on low-resolution copy number, gene expression and methylation ( Ben-David et al , 2018 ; Dancik et al , 2011 ; Domcke et al , 2013 ; Mohammad et al , 2019 ; Najgebauer et al , 2020 ; Peng et al , 2021 ; Salvadores et al , 2020 ; Vincent et al , 2015 ; Warren et al , 2021 ; Yu et al , 2019 ; Zhao et al , 2017 ). In addition, CNpare offers the option of comparing copy number profiles from a more functional point of view by using copy number signatures.…”

“…Cell lines are often selected based on their tissue of origin. However, new approaches are available that facilitate appropriate cell line selection based on molecular similarities such as gene expression, DNA methylation and genomics ( Ben-David et al , 2018 ; Dancik et al , 2011 ; Domcke et al , 2013 ; Mohammad et al , 2019 ; Najgebauer et al , 2020 ; Peng et al , 2021 ; Salvadores et al , 2020 ; Sinha et al , 2015 ; Vincent et al , 2015 ; Warren et al , 2021 ; Yu et al , 2019 ; Zhao et al , 2017 ). A subset of these approaches perform DNA copy number-based comparison at different resolutions including gene-level copy number, chromosome arm copy number, ploidy or genome doubling status ( Ben-David et al , 2018 ; Domcke et al , 2013 ; Najgebauer et al , 2020 ; Zhao et al , 2017 ).…”

Computational estimation of quality and clinical relevance of cancer cell lines

“…More importantly, only 9 of 22 studies address multi‐omic sources with disparate strategies for their integration. Here, we only focus on studies comparing tumours to CCLs, but we nevertheless note that 2 of the listed publications (Liu et al , 2019b ; Peng et al , 2021 ) extend their methods to more complex models such as tumoroids and PDx, highlighting different representative performances and quality across model complexities.…”

Three-dimensional organoid culture unveils resistance to clinical therapies in adult and pediatric glioblastoma

“…Many therapeutics currently in use are highly effective in cultured samples, yet have an incomplete response in human patients and inevitably lead to resistance (reviewed in [ 12 , 52 ]. Indeed, many cultured cancer lines have poor transcriptional fidelity to their clinical tumor counterparts, and this is exemplified by cultured brain tumor models where no low grade glioma and only ∼5% of GBM models can be correctly classified as brain tumors [34] . The use of single-cell sequencing analysis, genetically engineered mouse models, or tumor organoid models dramatically improved fidelity to clinical disease, with all tested GBM organoid models being transcriptionally classified as GBM with high confidence [34] .…”