Built by Rileen Sinha, Augustin Luna, Nikolaus Schultz, and Chris Sander using R Shiny.
Cell lines derived from human tumors are often used in pre-clinical cancer research, but some cell lines may be too different from tumors to be good models. Genomic and molecular profiles can be used to guide the choice of cell lines suitable for particular investigations, but not all features may be equally relevant. We present TumorComparer, a computational method for comparing cellular profiles with the flexibility to place a higher weight on functional features of interest. In a first pan-cancer application, we compare 600 cell lines and 8323 tumors of 26 cancer types, using weights to emphasize recurrent genomic alterations or expression dysregulation. We characterize the similarity of cell lines and tumors within and across cancers, identifying outlier and mislabelled cell lines as well as good matches, and identify cancers with an unusually high number of good or poor representative cell lines. The weighted similarity method in the future may be useful to assess genomic-molecular patient profiles for stratification in clinical trials and personalized choice of therapy.
The “Pre-Computed” tab allows users to explore the results of our systematic analysis available in the publication, while the “User Analysis” section provides users the option of running TumorComparer with their own data.
All cancer types from the publication are available. Balloon plots show the most relevant cell lines and a corresponding searchable table.
This tab allows users to run TumorComparer on their own data. Users will upload a zip file containing the input files with a pre-specified naming convention.
NOTE: For large computational pipelines or large amounts of data, it is suggested that users run TumorComparer locally.
Users are should review the publication for more information about these parameters. The default weights were used as part of the systematic analysis.
Users should rely on their understanding of the problem they are trying to address to select weights. Users can use 1 and 0 as a starting point for known cancer gene and default weights if the relative difference in the magnitude in importance between the sets of genes is unknown.
NOTE: You do NOT have to provide input files for all data types (expression: exp, mutation: mut, copy number: cna). Make sure either ALL files for a data type are included OR NONE of them are included, based on your available data.
Below is example data for the various supported data types.
Below is an example of GISTIC discretized
The TumorComparer software package contains a number of additional parameters that might be of interest to users.
We appreciate any feedback/suggestions you may have; please send feedback to publication corresponding authors.