Identify Interferon-associated LVs in PLIER models

J. Taroni 2018

The SLE WB compendium contains two data sets that examine interferon (IFN) modulating treatments. For context, IFNs are signaling molecules that modulate the immune response and there are multiple types:

• Type I - IFN-alpha, IFN-beta
• Type II - IFN-gamma
• Type III which are less well-characterized than I & II

We’re concerned with types I and II because these are the types we expect to be modulated by the treatments included in the data we’re working with (see more information in a data preparation notebook). It’s also worth noting that it can be difficult to distinguish between the two when it comes to gene expression data, i.e., asking which interferon is inducing changes in gene expression can be tricky.

In this notebook, we’ll identify latent variables in the SLE WB and recount2 PLIER models that are significantly associated with IFN-related gene sets.

Pipe and directory setup

`%>%` <- dplyr::`%>%`

# plot and result directory setup for this notebook
plot.dir <- file.path("plots", "08")
dir.create(plot.dir, recursive = TRUE, showWarnings = FALSE)
results.dir <- file.path("results", "08")
dir.create(results.dir, recursive = TRUE, showWarnings = FALSE)

PLIER trained on SLE WB compendium

sle.plier.results <- readRDS(file.path("results", "05", 
                                       "SLE-WB_PLIER_model.RDS"))

# get summary data.frame and write to file
sle.summary <- sle.plier.results$summary
readr::write_tsv(sle.summary, 
                 path = file.path("results", "05",
                                  "SLE-WB_PLIER_summary.tsv"))

sle.summary %>% 
  dplyr::filter(grepl("INTERFERON", pathway),
                FDR < 0.05)

Note that IFN gamma signaling (type II IFN) does not appear to be captured (at least at FDR < 0.05) with this model. Let’s take a look at the other gene sets significantly associated with these LVs.

sle.summary %>%
  dplyr::filter(`LV index` %in% c(6, 69, 110),
                FDR < 0.05)

sle.b.df <- reshape2::melt(sle.plier.results$B)
colnames(sle.b.df) <- c("Annotated", "Sample", "Value")
sle.b.df <- dplyr::mutate(sle.b.df,
                          LV = rep(paste0("LV", 1:nrow(sle.plier.results$B)),
                                   ncol(sle.plier.results$B))) %>%
              dplyr::select(c("Sample", "LV", "Annotated", "Value"))
readr::write_tsv(sle.b.df, 
                 path = file.path("results", "05",
                                  "SLE-WB_PLIER_B_tidy.tsv"))

PLIER trained on recount2

# load recount2 PLIER model
recount.plier <- readRDS(file.path("data", "recount2_PLIER_data", 
                                   "recount_PLIER_model.RDS"))

# get summary data.frame and write to file
recount.summary <- recount.plier$summary
readr::write_tsv(recount.summary, 
                 path = file.path(results.dir,
                                  "recount2_PLIER_summary.tsv"))

recount.summary %>%
  dplyr::filter(grepl("INTERFERON", pathway),
                FDR < 0.05) %>%
  dplyr::arrange(pathway, dplyr::desc(AUC))

The model trained on recount2 captures both IFN-alpha/beta signaling and IFN-gamma signaling. Let’s see what other pathways are associated with LV25, LV34, LV116, LV140, LV749, and LV827.

recount.summary %>%
  dplyr::filter(`LV index` %in% c(25, 34, 116, 140, 749, 827),
                FDR < 0.05) %>%
  dplyr::arrange(dplyr::desc(AUC))

We’ll investigate LV116 and LV140 further.