Implement Cellpanelr-like biomarker identification features in CanDI #55
Description
Implement features in CanDI similar to those in Cellpanelr, a tool designed to identify predictive biomarkers from cell line panel response data. Cellpanelr is an R package + an all-in-one Shiny app for data handling and visualization. It integrates cell line annotations, gene expression, and mutation data sets from DepMap and dose-response data from GDSC. User-provided cell line names are mapped to unique DepMap IDs, enabling the merging of user-provided response values with gene expression, mutation, and annotation data per cell line.
Key features:
- For each of the genes, Spearman’s correlation is calculated between cell line response and gene expression (log2[TPM + 1]).
- For each of 19,537 genes, a Mann-Whitney U test is performed to assess whether cell lines with a mutation in the gene respond differently than wild-type cell lines. Statistical significance is determined using the Benjamini-Hochberg method (FDR 0.05).
Key tasks:
- Review the Cellpanelr paper and extract relevant methods and features.
- Design and implement modules in CanDI for biomarker identification from cell panel response data, including data mapping, correlation/statistical analysis, and visualization tools.
- Ensure support for data integration (DepMap, GDSC), statistical association analyses (Spearman, Mann-Whitney U, Benjamini-Hochberg correction), and interactive/result visualization.
- Document all new features and provide usage examples.
Reference:
bioRxiv: https://www.biorxiv.org/content/10.1101/2022.11.02.514913v1.full
GitHub: https://github.com/dwassarman/cellpanelr
ShinyApp: https://dwassarman.shinyapps.io/cellpanelr/