Drug Response Modeling and Biomarker Discovery

Allows for building drug response models using screening data between bulk RNA-Seq and a drug response metric and two additional tools for biomarker discovery that have been developed by the Huang Laboratory at University of Minnesota. There are 3 main functions within this package. (1) calcPhenotype() is used to build drug response models on RNA-Seq data and impute them on any other RNA-Seq dataset given to the model. (2) GLDS() is used to calculate the general level of drug sensitivity, which can improve biomarker discovery. (3) IDWAS() can take the results from calcPhenotype() and link the imputed response back to available genomic (mutation and CNV alterations) to identify biomarkers. Each of these functions comes from a paper from the Huang research laboratory. Below gives the relevant paper for each function. The package is described in Maeser et al. (2021) "oncoPredict: an R package for predicting in vivo or cancer patient drug response and biomarkers from cell line screening data" . calcPhenotype() - Geeleher et al, Clinical drug response can be predicted using baseline gene expression levels and in vitro drug sensitivity in cell lines. GLDS() - Geeleher et al, Cancer biomarker discovery is improved by accounting for variability in general levels of drug sensitivity in pre-clinical models. IDWAS() - Geeleher et al, Discovering novel pharmacogenomic biomarkers by imputing drug response in cancer patients from large genomics studies.

oncoPredict

(Predict Response from Expression Data and Identify Cell line/Clinical Targets and Trends)

Additional details about this package can be found in our publication oncoPredict: an R package for predicting in vivo or cancer patient drug response and biomarkers from cell line screening data

An R package for drug response prediction and drug-gene association prediction. The prepared GDSC and CTRP matrices for the calcPhenotype() are located in the oncoPredict OSF.

• For drug response prediction, use calcPhenotype.
• For pre-clinical biomarker discovery, use GLDS.
• For clinical biomarker discovery, use IDWAS (for CNV or somatic mutation association with drug response) or indicate cc=TRUE (for gene expression association with drug response) in calcPhenotype().
• The link to updated CCLE gene expression data is found at depmap. We provide GDSC1/GDSC2 pre-processed expression and response data, as well as CTRP response data and depmap's CCLE expression data (18Q2) here.

User Notes:

• Reminder that it is best to logtransform the training and test datasets (i.e. log(TPM + 1)) before batch correction ("eb"). That is, use log-transformed expression data when using calcPhenotype.

R

• This directory contains all the R functions included in this package.

vignettes

• This directory contains vignettes which display detailed examples of the functionalities available in this package.

• IDWAS This directory contains examples of IDWAS code application for clinical drug-gene association prediction.

  • cnv.Rmd Example as to how to download CNV (copy number variation) data from the GDC database, then apply map_cnv() and idwas().
  • mut.Rmd Example as to how to download stomatic mutation data from the GDC database, then apply idwas().

• GLDS This directory contains examples of GLDS code application for pre-clinical drug-gene association prediction.

  • glds_GDSC.Rmd Example of GLDS application to GDSC data.

• calcPhenotype.Rmd Example of calcPhenotype() application.

man

• This directory contains .Rd (R documentation) files for each function. These files were automatically generated upon creation of the package.

NAMESPACE

• This file lists the functions to be imported and exported from this package.

DESCRIPTION

• This file contains the description documentation and metadata for this package.
• Dependencies and packages recommended for oncoPredict are listed here.

Figure 1.

Flowchart displaying the 3 primary functionalities available through oncoPredict (calcPhenotype, GLDS, IDWAS) as well as the files generated from each function and parameters. Functions and files generated are bold.