Bio+MedVis Challenge @ IEEE VIS 2023
Bio+MedVis Challenge Program 21 Sep 2023
The Bio+MedVis Challenge program has been announced. Please visit the program page for more information.
Submission Deadline Extended 4 Aug 2023
Based on your requests, we extended the submission deadline by one week to August 17. Check all important dates below.
This year’s challenge will be based on a study recently published in Cancer Cell by the ProCan® team (Gonçalves et al., 2022). The study generated a comprehensive pan-cancer proteomic map of 949 human cancer cell lines to aid in the discovery of cancer biomarkers and targets for the development of new cancer treatments. The primary goals of this challenge can be found in the “Challenge tasks” section. This year, in addition to the description below and the slack channel, we will also hold an online Q&A session where you can directly ask questions to the ProCan team. The date/time of this session will be announced here, as well as our slack (https://biovis2016.slack.com/) and twitter (@biovis_net) channels.
Biological Background and Data Description
The ProCan team used mass spectrometry-based proteomics to identify and quantify proteins in 949 human cell lines that represented 28 different tissue types and over 40 genetically and histologically diverse cancer types. These cell lines were treated with 625 anti-cancer compounds and for many of them over 17,000 gene essentialities had been tested via CRISPR-Cas9 knockout screens. The analysis resulted in a dataset of approximately 8,500 proteins that were quantified across the cell lines, capturing cell type features and evidence of post-transcriptional control of protein levels. This dataset was then further analyzed using a deep learning-based pipeline that integrated multi-omics, drug response, and CRISPR-Cas9 gene edits; this analysis revealed thousands of statistically significant protein biomarkers of cancer vulnerabilities that were not significant at the transcript level. Overall, the ProCan team found that the power of the proteome to predict drug response was very similar to that of the transcriptome. Interestingly, the team also found that random downsampling to only 1,500 proteins had limited impact on the ability to predict drug responses, consistent with protein networks being highly connected and co-regulated. This proteomic dataset is a high-quality resource for mechanistic investigation of network organisation and regulatory principles of the proteome, as well as for translational discoveries.
The data are provided by domain experts from ProCan, Children’s Medical Research Institute, The University of Sydney.
- A peptide data matrix of 949 cell lines with 79,621 quantified peptides.
- A protein data matrix of 949 cell lines with 8,498 quantified proteins.
- An annotation file of tissue and cancer types.
- Drug response data.
The intensity values of peptides and proteins have been averaged over six replicates. The drug response data record 578,238 half-maximal inhibitory concentrations (IC50) of the 625 anti-cancer drugs. The data are available under the following link: Dropbox
The figure (Gonçalves et al., 2022) below provides an overview of the study, which displays a comprehensive pan-cancer proteomic map. This map quantifies proteomes of 949 human cancer cell lines originating from 28 distinct tissues and over 40 diverse cancer types, analysed using six mass spectrometers.
The following tasks provide guidelines for your submission. We also welcome submissions that do not solve all of these tasks and only focus on a subset thereof:
- The first task is to create an easy-to-understand visualization of peptide and protein intensities by including phenotypic information about tissue and cancer types. This challenge requires a user-friendly and interactive solution that helps researchers explore the complex data in an engaging way and reveal hidden patterns. Here, the participants can define subcategories for protein groups, such as house-keeping proteins, tissue-specific and cancer-specific proteins.
- Extracting meaningful insights from the existing peptide and protein data matrices and annotation files, as the relevant information is not easily accessible or comprehensible in a tabular format. This challenge calls for participants to create an intuitive visualization that seamlessly connects drug responses to their respective proteins and cell lines.
- Create an interactive tool to visualise drug response data across cell lines in relation to their proteomic data. The tool should allow users to select a protein and analyse drugs that target them. Similarly, it should also allow users to select a specific drug and visualise proteins that are strongly associated with the drug response. The drug response data can be retrieved from https://www.cancerrxgene.org/downloads/bulk_download and protein-drug association information can be found in Supplementary Table 5 of the paper (Gonçalves et al., 2022).
Design challenge: Re-design an existing visualization
In addition to the main challenge, we also host a complementary design challenge. Teams that submitted entries to the main challenges are allowed to also submit to the design challenge track and vice versa.
For this challenge, submitters are asked to improve the current protein network plot in Fig. 7F (Gonçalves et al., 2022) by creating an interactive and responsive visualization of the protein network. This solution should allow users to zoom in and out, focus on specific proteins, and explore the relationships between them more easily. Suggestions for features that can be developed: pop-up annotations when a user hovers over a protein node; and a search function that enables users to quickly locate proteins of interest. Color-coding and distinct symbols can be used to represent different types of proteins or protein functions.
The Bio+MedVis Challenge, jointly organized by the BioVis and VCBM communities, will be organized as a half-day event within the IEEE VIS 2023 conference. The program will consist of invited talks, presentations of accepted submissions, and other supplementary sessions.
See the program for more details.
- Camera-ready version:
Submissions will be considered for talk and/or poster presentations. You will be requested to submit a two-page PDF abstract with up to 5 additional figures and a draft of your proposed poster (max 10MB) to the PCS new.precisionconference.com/submissions. If you want to submit more supplementary figures, please contact us. Last year, the acceptance rate for presentations was 89%; such an acceptance rate is not guaranteed. Selected submissions will be invited for talk presentations during the Bio+MedVis Challenge session within the IEEE VIS 2023 conference and also will be presented as posters.
Evaluation of Submissions
All submissions will be thoroughly reviewed by at least two reviewers, coming from the challenge chairs and members of the ProCan team. All accepted submissions will be published in the conference proceedings.
Please feel free to ask questions on our Slack channel if you need more
information and details about the data or tasks:
biovis2016.slack.com » #biovis-challenge2023
Otherwise, feel free to contact us at: email@example.com
The chairs of the Bio+MedVis Challenge @ IEEE VIS 2023
Barbora Kozlikova, Masaryk University, Czech Republic
Daniel Jönsson, Linköping University, Sweden
Renata Raidou, TU Wien, Austria
Sean O’Donoghue, Garvan Institute of Medical Research, Sydney
E. Gonçalves, et al., Pan-cancer proteomic map of 949 human cell lines, Cancer Cell. 40 (2022) 835–849.