This package is a user friendly implementation of the CASTLE (Concentration and Assay Specific Tumor Load Estimator) algorithm, as it is described in the accompanying scientific paper. Using the right analysis tool for your ddPCR analysis results when detecting low frequency variants can mean night and day for the performance of your pipeline. In this package functions for training a model and testing samples for low frequency variants are provided, as well as functionality to handle output .csv-files generated by QuantaSoft from Bio-Rad Laboratories Inc.

You can install the R -package from GitHub with:

# install.packages("devtools")
devtools::install_github("simondrue/castle")

Basic usage of castle consists of the following classic workflow:

# Load package
library(castle)

# Load training samples
training_data_path <- c("path/to/training/csv_files/")
training_samples <- import_QS_files(training_data_path)

# Train model
trained_model <- train_integrated_ddpcr_model(
  background_samples = training_samples
)

# Test some samples
test_samples <- c("path/to/test/csv_files/")
test_res <- test_tumor_sample_integrated(
  test_samples = test_samples,
  integrated_model = trained_model
)

To see a full example of how to do this with real world data, please see vignette("castle") which goes through a full analysis of a real world dataset.