scGPS introduction (original) (raw)
Here we look at one example use case to find relationship between clusters within one sample or between two sample
#run prediction for 3 clusters
cluster_mixedpop1 <- colData(mixedpop1)[,1]
cluster_mixedpop2 <- colData(mixedpop2)[,1]
#cluster_mixedpop2 <- as.numeric(as.vector(colData(mixedpop2)[,1]))
c_selectID <- 1
#top 200 gene markers distinguishing cluster 1
genes = DEgenes$id[1:200]
LSOLDA_dat1 <- bootstrap_prediction(nboots = 2, mixedpop1 = mixedpop2,
mixedpop2 = mixedpop2, genes=genes, c_selectID,
listData =list(),
cluster_mixedpop1 = cluster_mixedpop2,
cluster_mixedpop2 = cluster_mixedpop2)
c_selectID <- 2
genes = DEgenes$id[1:200]
LSOLDA_dat2 <- bootstrap_prediction(nboots = 2,mixedpop1 = mixedpop2,
mixedpop2 = mixedpop2, genes=genes, c_selectID,
listData =list(),
cluster_mixedpop1 = cluster_mixedpop2,
cluster_mixedpop2 = cluster_mixedpop2)
c_selectID <- 3
genes = DEgenes$id[1:200]
LSOLDA_dat3 <- bootstrap_prediction(nboots = 2,mixedpop1 = mixedpop2,
mixedpop2 = mixedpop2, genes=genes, c_selectID,
listData =list(),
cluster_mixedpop1 = cluster_mixedpop2,
cluster_mixedpop2 = cluster_mixedpop2)
c_selectID <- 4
genes = DEgenes$id[1:200]
LSOLDA_dat4 <- bootstrap_prediction(nboots = 2,mixedpop1 = mixedpop2,
mixedpop2 = mixedpop2, genes=genes, c_selectID,
listData =list(),
cluster_mixedpop1 = cluster_mixedpop2,
cluster_mixedpop2 = cluster_mixedpop2)
#prepare table input for sankey plot
LASSO_C1S2 <- reformat_LASSO(c_selectID=1, mp_selectID = 2,
LSOLDA_dat=LSOLDA_dat1,
nPredSubpop = length(unique(colData(mixedpop2)
[,1])),
Nodes_group ="#7570b3")
LASSO_C2S2 <- reformat_LASSO(c_selectID=2, mp_selectID =2,
LSOLDA_dat=LSOLDA_dat2,
nPredSubpop = length(unique(colData(mixedpop2)
[,1])),
Nodes_group ="#1b9e77")
LASSO_C3S2 <- reformat_LASSO(c_selectID=3, mp_selectID =2,
LSOLDA_dat=LSOLDA_dat3,
nPredSubpop = length(unique(colData(mixedpop2)
[,1])),
Nodes_group ="#e7298a")
LASSO_C4S2 <- reformat_LASSO(c_selectID=4, mp_selectID =2,
LSOLDA_dat=LSOLDA_dat4,
nPredSubpop = length(unique(colData(mixedpop2)
[,1])),
Nodes_group ="#00FFFF")
combined <- rbind(LASSO_C1S2,LASSO_C2S2,LASSO_C3S2, LASSO_C4S2 )
combined <- combined[is.na(combined$Value) != TRUE,]
nboots = 2
#links: source, target, value
#source: node, nodegroup
combined_D3obj <-list(Nodes=combined[,(nboots+3):(nboots+4)],
Links=combined[,c((nboots+2):(nboots+1),ncol(combined))])
library(networkD3)
Node_source <- as.vector(sort(unique(combined_D3obj$Links$Source)))
Node_target <- as.vector(sort(unique(combined_D3obj$Links$Target)))
Node_all <-unique(c(Node_source, Node_target))
#assign IDs for Source (start from 0)
Source <-combined_D3obj$Links$Source
Target <- combined_D3obj$Links$Target
for(i in 1:length(Node_all)){
Source[Source==Node_all[i]] <-i-1
Target[Target==Node_all[i]] <-i-1
}
#
combined_D3obj$Links$Source <- as.numeric(Source)
combined_D3obj$Links$Target <- as.numeric(Target)
combined_D3obj$Links$LinkColor <- combined$NodeGroup
#prepare node info
node_df <-data.frame(Node=Node_all)
node_df$id <-as.numeric(c(0, 1:(length(Node_all)-1)))
suppressMessages(library(dplyr))
Color <- combined %>% count(Node, color=NodeGroup) %>% select(2)
node_df$color <- Color$color
suppressMessages(library(networkD3))
p1<-sankeyNetwork(Links =combined_D3obj$Links, Nodes = node_df,
Value = "Value", NodeGroup ="color", LinkGroup = "LinkColor",
NodeID="Node", Source="Source", Target="Target", fontSize = 22)
p1
#saveNetwork(p1, file = paste0(path,'Subpopulation_Net.html'))