BCPLS CNA Hub Adjacency Analysis

Motivation

Some of the reported CNA hubs from the network analysis pipeline have very short genomic distances between themselves (i.e. ‘adjacent’) and ought to be considered for merging them into one CNA hub. By computing their Pearson correlation and enumerating common edges between adjacent CNA hubs, we can have a better idea on whether merging is appropriate. If two adjacent CNA hubs are highly correlated, it is very likely that these CNA hubs should have been merged into one CNA hub by the fixed-order clustering algorithm (Pei Wang, 2005), but were not merged due to a few outlying values (large sample discrepancies between two adjacent CNA). Additionally, knowing how large the CNA hub intervals are and how much overlap there is in edges helps determine whether one should consider merging the adjacent CNA hubs into one hub and bundling their corresponding trans edges. In what follows, we consider these factors for merging adjacent hubs on a case-by-case basis for CNA–protein and CNA–mRNA breast cancer networks, respectively.

Define functions to compute correlation between adjacent CNA hubs and to report the number of overlapping edges between them.

check_adjacent_correlation <- function(hub, show_plot = T, ecna = NULL) { 

    res <- round(cor(ecna[,hub[1]], ecna[,hub[2]]),2)
  #if (show_plot & !is.null(ecna)) { 
  library(ggplot2)
  gg <- ggplot2::qplot(x = ecna[,hub[1]], y  = ecna[,hub[2]],
                 ylab = names(hub)[1],
                 xlab = names(hub)[2]) + theme_bw() + 
    annotate(geom="text", x=-1, y=1, label=paste("Correlation", 
                                                 res),
             color="red")  
  #}
  #res
  gg
}
num_common_hub_edges <- function(hub, ig) { 
  huba <- as_ids(E(ig)[inc(V(ig)$name %in% hub[1])])
  hubb <- as_ids(E(ig)[inc(V(ig)$name %in% hub[2])])
  a <- sapply(strsplit(x = huba, split = "|", fixed = TRUE), function(x) x[2])
  b <- sapply(strsplit(x = hubb, split = "|", fixed = TRUE), function(x) x[2])
  length(intersect(a, b))
}

Adjacent BCPLS CNA hubs (Protein)

Load the protein and CNA data.

suppressPackageStartupMessages(library(Biobase))
protset <- readRDS(file = "data/prot-expression-set.rds")
y <- t(exprs(protset))
cnaset <- readRDS(file = "data/cna-expression-set.rds")
x <- t(exprs(cnaset))

Load spaceMap’s CNA–protein network.

pgraphml <- "/Users/bioinformatics/repos/neta-bcpls/neta/spacemap-prot-boot-vote.graphml"
stopifnot(file.exists(pgraphml))
suppressPackageStartupMessages(library(igraph))
pig <- read_graph(file = pgraphml, format = "graphml")

hubs <- c("16_68039840_68051376_0","16_68063383_71366920_0")
names(hubs) <- c("16q22.1-22.2 (68-71 Mb)", "16q22.1 (68-68 Mb)")

check_adjacent_correlation(hub = hubs, ecna = x )

hubs <-  c("17_20255440_20890727_0", "17_19777544_19864430_0")
names(hubs) <- c("17p11.2 (20-21 Mb)", "17p11.2 (20-20 Mb)")

check_adjacent_correlation(hub = hubs, ecna = x )

Adjacent BCPLS CNA hubs (RNA)

Load spaceMap’s CNA–mRNA network.

rgraphml <- "/Users/bioinformatics/repos/neta-bcpls/neta/spacemap-rna-boot-vote.graphml"
stopifnot(file.exists(rgraphml))
library(igraph)
rig <- read_graph(file = rgraphml, format = "graphml")

hubs <-  c("5_88348122_89201306_0","5_89203610_90774367_0")
names(hubs) <- c("5q14.3 (88-89 Mb)","5q14.3 (89-91 Mb)")

check_adjacent_correlation(hub = hubs, ecna = x )

hubs <-  c("5_164621218_166461162_0","5_166465008_168130127_0")
names(hubs) <- c("5q34 (160-170 Mb)","5q34 (170-170 Mb)")

check_adjacent_correlation(hub = hubs, ecna = x )

hubs <-  c("10_424200_4026124_0","10_4028083_11642478_0")
names(hubs) <- c("10p15.1-15.3 (0.42-4 Mb)","10p14-15.1 (4-12 Mb)")

check_adjacent_correlation(hub = hubs, ecna = x)

hubs <-  c("9_9411107_15241495_0","9_15259791_20352615_0")
names(hubs) <- c("9p22.3-23 (9.4-15 Mb)","9p21.3-22.3 (15-20 Mb)")

check_adjacent_correlation(hub = hubs, ecna = x)