Secreted protein signaling activity change between two phenotypes

This tutorial demonstrates how to infer the activity change of secreted proteins between two phenotypes, such as treatment vs control, tumor vs normal, or responder vs non-responder. The input expression values, from either RNA sequencing or MicroArray, should be transformed by log2(x+1). Of note, x could be FPKM, RPKM, or TPM for RNA-Seq data.

For this demonstration, we employ a clinical cohort of patients with discoid lupus erythematosus, who received a single dose of anti-IFNG antibody AMG 811. Using the SecAct framework, we will predict the activity change of secreted proteins between treatment and control groups.

Prepare expression data

This section shows how to read an expression matrix to the R environment and separate samples into the treatment and control groups.

library(SecAct)

# prepare expression matrix
dataPath <- file.path(system.file(package="SecAct"), "extdata/")
expr <- read.csv(paste0(dataPath,"GSE100093.IFNG.expr.gz"), sep="\t")

# read meta infomation
meta <- read.csv(paste0(dataPath,"GSE100093.IFNG.meta"), sep="\t")

# subset treatment and control groups
expr.treatment <- expr[, meta[,"Treatment"]=="Anti-IFNG"]
expr.control <- expr[, meta[,"Treatment"]=="Control"]

Infer activity change

We input expr.treatment and expr.control into SecAct.activity.inference to infer the activity change of secreted proteins between treatment and control groups. SecAct calculates the OVERALL activity change between two phenotypes with default setting is.singleSampleLevel = FALSE. You can also obtain the activity change for each SINGLE sample in expr.treatment by setting is.singleSampleLevel = TRUE. Of note, if your samples are paired, just change the default setting is.paired = FALSE to TRUE.

# infer activity change; ~2 mins
res <- SecAct.activity.inference(
  inputProfile = expr.treatment, 
  inputProfile_control = expr.control,
  is.singleSampleLevel = FALSE,
  is.paired = FALSE
)

# res$zscore stores activity change
act.change <- res$zscore

# show activity change
head(act.change)

A positive value means the secreted protein has high activity in the treatment group while a negative value means the secreted protein has high activity in the control group.

# show IFNG activity change
act.change["IFNG","Change"]

We can see that IFNG activity was reduced as expected, consistent with the treatment group receiving anti-IFNG therapy.

Visualize activity change

User can visualize any secreted proteins of interest using SecAct.bar.plot. Here, we select the top up- and down-regulated secreted proteins (SPs) between treatment and control as examples.

# select the top secreted proteins
n <- 10
up.SPs <- names(sort(act.change[,"Change"],decreasing=T))[1:n]
down.SPs <- names(sort(act.change[,"Change"]))[1:n]

SPs <- c(up.SPs, down.SPs)
# user can assign any secreted proteins to SPs

# generate a vector
fg.vec <- act.change[SPs,"Change"]

# visualize in bar plot
SecAct.bar.plot(fg.vec, title="Activity Change (treatment vs control)")

A positive value means the secreted protein has high activity in the treatment group while a negative value means the secreted protein has high activity in the control group.

Run a differential profile

If you already have a differential profile (or signature) instead of the expression matrix from two phenotypes, you can directly run SecAct.activity.inference with is.differential = TRUE.

# compare two phenotypes to generate a differential profile
expr.diff <- matrix(
  data = rowMeans(expr.treatment) - rowMeans(expr.control), 
  ncol = 1,
  dimnames = list(
    row_names = rownames(expr.treatment),
    col_names = "Diff"
  )
)

# Run a differential profile; ~2 mins
res <- SecAct.activity.inference(
  inputProfile = expr.diff, 
  is.differential = TRUE
)

# res$zscore stores activity change
act.change <- res$zscore

# show activity change
head(act.change)