Clinical drug response prediction from cell line screens, using non- linear alignment of tumor and cell line gene expression profiles.
Linear alignment of gene expression profiles by systematic comparison of principal components.
Weighted orthogonal non-negative (WON) parallel factor analysis (PARAFAC)
Performs gene-set enrichment analysis with sample permutation. Very flexible, including support for RNAseq data.
A mixed model approach to IC50 estimation that enables simultaneous estimation of IC50 values across the entire set of cell lines and compounds. We show that this approach improves the accuracy of the estimates and significantly reduces the compute time.
A two-stage penalized linear regression approach that uses upstream (genomics) and downstream (transcriptomics) to predict drug response. It results in models that are more interpretable while maintaining similar predictive performance.
R Package can be installed from CRAN install.packages("TANDEM")
Identifies co-occurrence and mutual exclusivity in somatic mutations using an elegant analytical null-model, which we show to faithfully recapitulate the nominal rates. The method suggests that many of the reported co-occurrences are in fact expected based on chance alone.
R and Python Packages
Prioritizes oncogenes and tumor suppressor genes based on the integration of various molecular data types.
TCGA pan-cancer results
Pinpoint driver genes in focal recurrent aberrations (across tumor samples) in DNA somatic copy number data.
A software package with samplers for Bayesian inference of computational models
Code and Documentation
R package for multivariate posterior distribution approximation from Monte Carlo samples.
See also our Github page: https://github.com/NKI-CCB