Plot drug response curves for combo data
Source:R/plot-combo-response_curve.R
plot_dose_response_combo.RdPlot drug response curves for combo data
Usage
plot_dose_response_combo(
dt_average,
drug1_name,
drug2_name,
cl_name,
normalization_type = "GR",
colors_vec = NULL,
split_by_conc = FALSE
)Arguments
- dt_average
data.table representing data from the
Averagedassay, outputted bygDRutils::convert_se_assay_to_dt(se, "Averaged")and comboSummarizedExperiment- 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 name to be plotted (identifiers
CellLineName)- normalization_type
string with normalization_types to be selected one of: "GR" ("GRvalue") or "RV" ("RelativeViability")
- colors_vec
character vector with colors for
group_names- name or hex value note that the first color will be assigned to the min value ofConcentration_2, and the last one - to the max ofConcentration_2; the default is the orange-black palette- split_by_conc
split_by_conc logical flag indicating whether curves for
Concentration_2should be plotted on a single plot or separately
Examples
mae <- gDRutils::get_synthetic_data("combo_matrix")
se <- mae[[gDRutils::get_supported_experiments("combo")]]
dt_average <- gDRutils::convert_se_assay_to_dt(se, "Averaged")
cl_name <- "cellline_BC"
drug1_name <- "drug_011"
drug2_name <- "drug_021"
plot_dose_response_combo(dt_average = dt_average,
drug1_name = drug1_name,
drug2_name = drug2_name,
cl_name = cl_name)
plot_dose_response_combo(dt_average = dt_average,
drug1_name = drug1_name,
drug2_name = drug2_name,
cl_name = cl_name,
split_by_conc = TRUE)
plot_dose_response_combo(dt_average = dt_average,
drug1_name = drug1_name,
drug2_name = drug2_name,
cl_name = cl_name,
normalization_type = "RV",
colors_vec = c("lightblue", "darkblue"))