Entropy Based Method for the Detection of Significant Variation in Gene Expression Data

An implementation of a method based on information theory devised for the identification of genes showing a significant variation of expression across multiple conditions. Given expression estimates from any number of RNA-Seq samples and conditions it identifies genes or transcripts with a significant variation of expression across all the conditions studied, together with the samples in which they are over- or under-expressed. It also detects genes whose relative isoform usage changes across samples (isoform switching). Zambelli et al. (2018) .

RNentropy

This is the implementation of a method based on information theory devised for the identification of genes showing a significant variation of expression across multiple conditions. Given expression estimates from any number of RNA-Seq samples and conditions it identifies genes or transcripts with a significant variation of expression across all the conditions studied, together with the samples in which they are over- or under-expressed. Zambelli F. et al. (2018).

A detailed walk-through on how to use RNentropy is available at Zambelli F., Pavesi G. (2021).

The main branch contains the 1.3.3 release. The exact CRAN 1.2.3 release sources are preserved by the v1.2.3 tag.

Installation

Install the stable release from CRAN:

install.packages("RNentropy")

Install version 1.3.3 from GitHub:

pak::pak("Federico77z/RNentropy")

Example

This is a basic example showing how to use RNentropy. Please see Zambelli F., Pavesi G. (2021) for more info.

library(RNentropy)
# basic example code
##load expression values and experiment design
data("RN_Brain_Example_tpm", "RN_Brain_Example_design")
#Run RNentropy 
Results <- RN_calc(RN_Brain_Example_tpm, RN_Brain_Example_design)
#select only genes with significant changes of expression
Results <- RN_select(Results)
#Compute the Point Mutual information Matrix
Results <- RN_pmi(Results)

Starting from an expression file

RNentropy() reads normalized expression values directly from a text file. The package includes a small synthetic example with three conditions and two replicates per condition. The first three columns contain transcript IDs, gene IDs, and comments; the remaining columns contain expression values.

Example_file <- system.file(
  "extdata", "RNentropy_example.tsv",
  package = "RNentropy"
)
Sample_names <- c("A_1", "A_2", "B_1", "B_2", "C_1", "C_2")
Example_design <- matrix(
  c(1, 0, 0,
    1, 0, 0,
    0, 1, 0,
    0, 1, 0,
    0, 0, 1,
    0, 0, 1),
  nrow = 6, byrow = TRUE,
  dimnames = list(Sample_names, c("A", "B", "C"))
)

File_Results <- RNentropy(
  Example_file,
  tr.col = 1,
  design = Example_design,
  header = FALSE,
  skip.col = c(2, 3),
  col.names = Sample_names
)
File_Results <- RN_select(File_Results)
File_Results$selected

Isoform-switch analysis

The package includes the complete historical S7 data used to validate the isoform-switch implementation. It contains expression values for 81,314 transcripts from 28,426 genes across six tissues. Load the data frame and run the analysis with RN_iso_calc():

data("RN_IsoSwitch_Example_S7")
S7_Results <- RN_iso_calc(
  RN_IsoSwitch_Example_S7,
  gene.col = "GENE_ID"
)
S7_Results <- RN_iso_select(S7_Results)
S7_Results$gene_status
S7_Results$sample_status
S7_Results$pv
S7_Results$gene_pv
S7_Results$selected

RN_iso_select() combines the finite sample p-values of each gene with status TESTED into one gene-level p-value with the Simes method, and retains genes whose gene-level p-value is at or below the threshold; no correction is applied across genes. Its gene_pv argument is expressed as a raw p-value and defaults to 0.05. The selected table reports the gene-level p-value (ISO_GENE_PV) and the uncorrected per-sample isoform-switch scores (ISO_PV_<sample>), which show in which samples the change occurs.

Starting from a transcript expression file

The same synthetic file can be analyzed for changes in relative isoform usage. Here the gene-ID column is retained, while the comments column is skipped.

Example_file <- system.file(
  "extdata", "RNentropy_example.tsv",
  package = "RNentropy"
)
Sample_names <- c("A_1", "A_2", "B_1", "B_2", "C_1", "C_2")
Example_design <- matrix(
  c(1, 0, 0,
    1, 0, 0,
    0, 1, 0,
    0, 1, 0,
    0, 0, 1,
    0, 0, 1),
  nrow = 6, byrow = TRUE,
  dimnames = list(Sample_names, c("A", "B", "C"))
)

Iso_File_Results <- RNentropy_iso_switch(
  Example_file,
  tr.col = 1,
  gene.col = 2,
  design = Example_design,
  header = FALSE,
  skip.col = 3,
  col.names = c("GENE_ID", Sample_names)
)
Iso_File_Results <- RN_iso_select(Iso_File_Results)
Iso_File_Results$gene_status
Iso_File_Results$selected

This compact dataset is synthetic and intended to demonstrate input handling and expected analysis behavior. The complete S7 dataset above provides the historical biological example. See ?RNentropy_iso_switch for all file-input options.

Independent C++ implementation

The independently developed C++ reference implementation, including the historical isoform-switch executable, is archived at Federico77z/RNentropy_cpp.