Skip to contents

Plot line plot of combination index

Usage

plot_combination_index(
  dt_excess,
  dt_isobolograms,
  drug1_name,
  drug2_name,
  cl_name,
  normalization_type = "GR",
  iso_levels = c("0.25", "0.5", "0.75"),
  colors_vec_iso = NULL
)

Arguments

dt_excess

data.table representing data from the excess assay, outputted by gDRutils::convert_se_assay_to_dt(se, "excess") and combo SummarizedExperiment

dt_isobolograms

data.table representing data from the isobolograms assay, outputted by gDRutils::convert_se_assay_to_dt(se, "isobolograms") and combo SummarizedExperiment

drug1_name

string with drug name to be plotted (identifiers DrugName)

drug2_name

string with co-drug name to be plotted (identifiers DrugName_2)

cl_name

string with cell line to be plotted (identifiers CellLineName)

normalization_type

string with normalization_types to be selected one of: "GR" ("GRvalue") or "RV" ("RelativeViability")

iso_levels

character vector with isobologram levels to be selected

colors_vec_iso

character vector of colors (valid name or hex) used for the isolines; the default is the dark red-orange palette

Value

ggplot object containing combination index plot at different ratios of the two drugs

Examples

cl_name <- "cellline_BC"
drug1_name <- "drug_001"
drug2_name <- "drug_026"

mae <- gDRutils::get_synthetic_data("combo_matrix")
se <- mae[[gDRutils::get_supported_experiments("combo")]]
dt_excess <- gDRutils::convert_se_assay_to_dt(se, "excess")
dt_isobolograms <- gDRutils::convert_se_assay_to_dt(se, "isobolograms")

plot_combination_index(dt_excess,
                       dt_isobolograms,
                       drug1_name, drug2_name,
                       cl_name,
                       normalization_type = "GR")


plot_combination_index(dt_excess,
                       dt_isobolograms,
                       drug1_name, drug2_name,
                       cl_name,
                       normalization_type = "RV",
                       colors_vec_iso = c("darkblue", "darkcyan"))


cl_name <- "cellline_JE"
drug1_name <- "drug_011"
drug2_name <- "drug_026"

plot_combination_index(dt_excess,
                       dt_isobolograms,
                       drug1_name, drug2_name,
                       cl_name,
                       normalization_type = "RV",
                       iso_levels = "0.5")