pgxR is a pharmacogenomics toolkit for fitting, visualizing, and comparing dose-response curves across cancer cell lines. It provides a complete workflow from raw viability data to publication-ready figures, with built-in support for population stratification by ancestry and genomic feature status.
- Fit 4-parameter log-logistic dose-response models via
drc - Extract IC50, AUC, and Hill slope in a single tidy table
- Visualize curves stratified by ancestry (HGDP+1KGP superpopulations)
- Facet by any genomic feature — translocation, driver mutation, CNV status
- Compare sensitivity across groups with built-in ANOVA and Wilcoxon tests
- Flexible column mapping
# install.packages("devtools")
devtools::install_github("iLOVESenescence/pgxR")library(pgxR)
# load built-in paclitaxel example data
data(pgxr_example)
# aggregate replicates
agg <- combineReps(pgxr_example)
# fit dose-response models
fits <- fitALL(agg, unique(agg$cell_line))
# generate smooth predictions
preds <- predictDRC(fits, agg)
# extract all metrics in one table
metrics <- summarizeDRC(fits, agg, preds)
#>
#> Estimated effective doses
#>
#> Estimate Std. Error Lower Upper
#> e:1:50 3.5364 0.3410 2.5897 4.4832
#>
#> Estimated effective doses
#>
#> Estimate Std. Error Lower Upper
#> e:1:50 152.766 40.390 40.626 264.906
#>
#> Estimated effective doses
#>
#> Estimate Std. Error Lower Upper
#> e:1:50 5.958059 0.095215 5.693701 6.222418
#>
#> Estimated effective doses
#>
#> Estimate Std. Error Lower Upper
#> e:1:50 19.5810 1.3001 15.9713 23.1907
#>
#> Estimated effective doses
#>
#> Estimate Std. Error Lower Upper
#> e:1:50 16.1960 1.7452 11.3506 21.0414
#>
#> Estimated effective doses
#>
#> Estimate Std. Error Lower Upper
#> e:1:50 82.9934 3.1928 74.1289 91.8579
#>
#> Estimated effective doses
#>
#> Estimate Std. Error Lower Upper
#> e:1:50 35.0818 1.5379 30.8120 39.3516
#>
#> Estimated effective doses
#>
#> Estimate Std. Error Lower Upper
#> e:1:50 42.6814 1.3537 38.9228 46.4399
metrics
#> cell_line ancestry feature IC50 IC50_lower IC50_upper AUC
#> 1 pLC1 AFR NF1-del 3.536447 2.589674 4.483219 0.5920703
#> 2 pLC2 AFR KRAS-mut 152.766177 40.625936 264.906418 0.1257630
#> 3 pLC3 EUR NF1-del 5.958059 5.693701 6.222418 0.5333453
#> 4 pLC4 EUR TP53-mut 19.580997 15.971253 23.190741 0.3234936
#> 5 pLC5 EAS TP53-mut 16.196019 11.350596 21.041442 0.3574289
#> 6 pLC6 EAS KRAS-mut 82.993387 74.128870 91.857905 0.1443219
#> 7 pLC7 AMR WT 35.081798 30.812033 39.351564 0.2848864
#> 8 pLC8 AMR WT 42.681382 38.922816 46.439947 0.2593990
#> hill_slope
#> 1 -2.238351
#> 2 -1.453203
#> 3 -3.041203
#> 4 -2.224502
#> 5 -2.247405
#> 6 -1.738232
#> 7 -2.061085
#> 8 -2.401667# dose-response curves by ancestry, faceted by genomic feature
plotDRCAnc(agg, preds, title = "Paclitaxel Response by Ancestry")
# IC50 comparison across ancestry groups with p-values
plotSensitivity(metrics, group_col = "ancestry", metric = "IC50",
title = "Paclitaxel IC50 by Ancestry")
pgxR expects columns dose, response, cell_line, ancestry, and
feature. If your CSV uses different names, use col_map:
raw <- loadData(
"my_data.csv",
col_map = list(feature = "translocation")
)
# if your replicates are tracked in a column other than the default,
# specify it with replicate_col
agg <- combineReps(raw, replicate_col = "experiment")
# standardize ancestry labels to HGDP+1KGP abbreviations
raw$ancestry <- standardizeAncestry(as.character(raw$ancestry))
GPL (>= 3)