delboy is an R package for conducting differential-expression analyses on RNA-seq data in which there are exactly two groups to be contrasted. The method is designed to improve sensitivity for under-powered data-sets, in which the effect sizes are small and there are few replicates, while controlling the False Discovery Rate (FDR).
delboy - Differential-representation analysis by Elastic-net Logistic regression with BinOmial-thinning validitY tests.
You can read about the method in the companion manuscript.
# install.packages("devtools")
devtools::install_github("alextkalinka/delboy")Input data should be a data frame of normalized counts in which there is a gene column with the remaining columns being sample columns.
db <- delboy::run_delboy(
data = expr_data_frame,
group_1 = c("ctrl-1","ctrl-2","ctrl-3"),
group_2 = c("treat-1","treat-2","treat-3"),
filter_cutoff = 40,
gene_column = "gene_id",
batches = NULL
)
# To print a summary report to the console:
db
# To extract a data frame of hits
# (includes a 'Predicted_False_Positive' column):
my_hits <- delboy::hits(db)
To plot validation performance next to original data showing the false-positive decision boundary (axes limits can be controlled using the xlim and ylim arguments):
plot(db, type = "lfc_expr")To visualize false negatives in the validation data relative to the false-positive decision boundary:
plot(db, type = "lfc_expr_FN")To plot the distrubution of log-fold changes used for the validation data:
plot(db, type = "lfc_nonnull")Kalinka, A. T. (2020). Improving the sensitivity of differential-expression analyses for under-powered RNA-seq experiments. bioRxiv 10.1101/2020.10.15.340737.
Please contact Alex Kalinka.