Article Text

Download PDFPDF

927 Streamlining cancer immunotherapy research with the CRI iAtlas data resource and web portal
  1. Vésteinn Thorsson1,
  2. Carolina Heimann1,
  3. Andrew Lamb2,
  4. David Gibbs1,
  5. Dante Bortone3,
  6. Sarah Dexheimer3,
  7. Steven Vensko3,
  8. Yooree Chae2,
  9. Ilya Shmulevich1,
  10. Benjamin Vincent3 and
  11. James Eddy2
  1. 1Institute for Systems Biology, Seattle, WA, United States
  2. 2Sage Bionetworks, Seattle, WA, United States
  3. 3University of North Carolina at Chapel Hill, Chapel Hill, NC, United States


Background With the increased volume of genomics data from studies involving treatment with immune checkpoint inhibition (ICI) and other immunotherapies, researchers remain unable to to make full use of results due to lack of comprehensive access to data or th ability to compare outcomes across datasets.The Cancer Research Institute (CRI) iAtlas1 ( is a comprehensive web platform for interactive data exploration and discovery in immuno-oncology, originating in a study by The Cancer Genome Atlas (TCGA).1–3 iAtlas provides topic-oriented analysis modules, each generating visualizations and statistics for studying interactions between tumors and the immune microenvironment (figure 1).

Methods Immunogenomic features from 15 ICI trials encompassing 1,142 samples were processed with a standardized bioinformatics workflow4 and incorporated into iAtlas, augmenting the 11,535 patient samples from TCGA1–3 and the Pan-Cancer Analysis of Whole Genomes5 consortia. A compendium of in-development immunotherapy drug targets6 and results of a study of germline genetic contribution to immune response in cancer7 were included. For efficient access, all data were incorporated into a relational database, and programmatic access was made available through an application programming interface (API) (figure 2). The set of available iAtlas modules was vastly extended, and numerous improvements were made to the codebase. Users can now define sample cohorts and sample groups based on any available categorical or numerical variable.

Results iAtlas provides 17 interactive analysis modules (table 1) to explore immune-cancer interactions, immunotherapy treatment, and outcomes in 12,677 patient samples. Six modules are dedicated to ICI studies: dataset overview, immune readouts, immunomodulators, clinical outcome, regression analysis, and a machine learning module to enable identification of factors associated with response to therapy (figure 3). We added modules to explore how germline variation and copy number alterations relate to immune response, and how receptor-ligand interactions mediate interactions among tumor and immune cells (figure 4). Docker images using Common Workflow Language descriptors are provided so that researchers can run iAtlas workflows on their own data. Computational notebooks are provided to illustrate and explain iAtlas code, plots, and functionality and to facilitate integration of iAtlas data with data sourced from a researcher’s own study.

Conclusions iAtlas serves as a repository and resource for harmonized data on immune response in cancer and response to immunotherapy. iAtlas enables researchers to readily test hypotheses and access data through multiple modalities: an interactive web portal, data download, tools,8 and computational workflows and notebooks.

Acknowledgements This work is supported by the Cancer Research Institute. We thank Allison Kudla, Institute for Systems Biology, for generating the illustration used in the Cell-Interaction Diagram module and for web design and implementation.


  1. Eddy JA, Thorsson V, Lamb AE, Gibbs DL, Heimann C, Yu JX, et al. CRI iAtlas: an interactive portal for immuno-oncology research. F1000Research 2020;9:1028.

  2. Hutter C, Zenklusen JC. The cancer genome atlas: creating lasting value beyond its data. Cell 2018;173:283–5.

  3. Thorsson V, Gibbs DL, Brown SD, Wolf D, Bortone DS, Ou Yang T-H, et al. The immune landscape of cancer. Immunity 2018;48:812–830.e14.

  4. Bortone DS, Vensko SP, Dexheimer S, Thorsson V, Zappasodi R, Rudqvist N-P, Vincent, BG et al. Generalizability of predictive versus prognostic indicators from published transcriptomic associations with tumor response to immune checkpoint inhibition. SITC Annual Meeting 2022, submitted.

  5. ICGC/TCGA Pan-Cancer Analysis of Whole Genomes Consortium. Pan-cancer analysis of whole genomes. Nature 2020;578:82–93.

  6. Upadhaya S, Hubbard-Lucey VM, Yu JX. Immuno-oncology drug development forges on despite COVID-19. Nat Rev Drug Discov 2020;19:751–2.

  7. Sayaman RW, Saad M, Thorsson V, Hu D, Hendrickx W, Roelands J, et al. Germline genetic contribution to the immune landscape of cancer. Immunity 2021;54:367–386.e8.

  8. Gibbs DL. Robust classification of immune subtypes in cancer. bioRxiv 2020:2020.01.17.910950.

Abstract 927 Figure 1

CRI iAtlas ExplorerEntry into exploration of immune response in cancer with iAtlas. Researchers start by defining cohorts and sample groups, and can then explore and visualize results using any of 17 analysis modules (left navigation bar, and bottom right of image).

Abstract 927 Figure 2

iAtlas 2.0 infrastructureThe infrastructure underlying the iAtlas web application (top right) has four main components: 1. R scripts to (i) download data files from Synapse; (ii) extract tables from files and perform any transformations to fit database schema; and (iii) build PostgreSQL database tables, which are stored in Aurora DB. 2. Python package that implements a GraphQL API service using Flask; GraphQL query requests are resolved to SQL database queries using SQLAlchemy. 3. R package that implements a GraphQL API client to retrieve data (and transform JSON responses into tables) within the app, or as part of a regular R session. 4. Shiny/R app code, including a set of reusable Shiny modules in the iatlas.modules R package; app is hosted using and deployed manually using the rsconnect package

Abstract 927 Figure 3

iAtlas ICI Machine Learning ModuleTo identify factors that may be associated with response to Immune Checkpoint Inhibition (ICI), users choose test and training sets, factors of interest, and the response variable (top). After selection of the modeling method (here, Random Forest – other choices are Elastic Net, Logistic regression and Gradient Boosting) and parameters, the model can be trained (bottom left) and statistics are reported after running on the test set(bottom right)

Abstract 927 Figure 4

Interactions in the Tumor Immune MicroenvironmentIn this iAtlas analysis module, Cell-Interaction Diagram, the estimated levels of cells and associated ligands and receptors that bind are shown within a selected group of samples, in this case ovarian (OV) tumor samples in the TCGA. Users can elect to show interactions superimposed on an illustration (left) or on a node-edge network diagram (right). A related module, Extracellular Networks, infers from data the likely ligand-receptor-mediated cellular interactions in the microenvironment and displays those as a node-edge network diagram.

Abstract 927 Table 1

Interactive analysis modules available in iAtlas

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.