Load Libraries

library(Seurat)
library(WGCNA)
library(cluster)
library(parallelDist)
library(ggsci)
library(emmeans)
library(lme4)
library(ggbeeswarm)
library(genefilter)
library(tidyverse)
library(reshape2)
library(igraph)
library(gProfileR)
library(ggpubr)
library(here)
library(ggforce)

Extract Cells for WGCNA

Calculate softpower

enableWGCNAThreads()

Allowing parallel execution with up to 79 working processes.

datExpr<-as.matrix(t(ventric[["SCT"]]@scale.data[ventric[["SCT"]]@var.features,]))
gsg = goodSamplesGenes(datExpr, verbose = 3);

 Flagging genes and samples with too many missing values...
  ..step 1

powers = c(c(1:10), seq(from = 12, to=40, by=2))
sft=pickSoftThreshold(datExpr,dataIsExpr = TRUE, powerVector = powers, corOptions = list(use = 'p'), 
                      networkType = "signed")

   Power SFT.R.sq  slope truncated.R.sq  mean.k. median.k.   max.k.
1      1    0.186 -56.60          0.593 2.50e+03  2.50e+03 2570.000
2      2    0.536 -45.50          0.561 1.26e+03  1.25e+03 1350.000
3      3    0.924 -36.60          0.906 6.32e+02  6.27e+02  722.000
4      4    0.958 -24.60          0.975 3.18e+02  3.15e+02  398.000
5      5    0.950 -16.50          0.970 1.61e+02  1.58e+02  228.000
6      6    0.945 -11.60          0.971 8.18e+01  7.94e+01  135.000
7      7    0.945  -8.70          0.964 4.17e+01  4.00e+01   84.200
8      8    0.948  -6.70          0.961 2.14e+01  2.01e+01   54.800
9      9    0.964  -5.33          0.970 1.10e+01  1.01e+01   37.200
10    10    0.962  -4.40          0.964 5.75e+00  5.12e+00   26.200
11    12    0.981  -3.17          0.979 1.62e+00  1.31e+00   14.400
12    14    0.983  -2.48          0.978 4.93e-01  3.35e-01    8.650
13    16    0.965  -2.09          0.962 1.67e-01  8.65e-02    5.510
14    18    0.966  -1.79          0.966 6.57e-02  2.24e-02    3.650
15    20    0.947  -1.62          0.949 3.01e-02  5.85e-03    2.490
16    22    0.396  -2.00          0.345 1.57e-02  1.54e-03    1.730
17    24    0.389  -1.84          0.343 9.11e-03  4.09e-04    1.220
18    26    0.381  -1.74          0.327 5.67e-03  1.09e-04    0.873
19    28    0.960  -1.24          0.952 3.71e-03  2.95e-05    0.632
20    30    0.942  -1.21          0.927 2.53e-03  8.08e-06    0.465
21    32    0.956  -1.18          0.944 1.79e-03  2.25e-06    0.390
22    34    0.336  -1.82          0.232 1.29e-03  6.33e-07    0.329
23    36    0.981  -1.14          0.976 9.62e-04  1.80e-07    0.278
24    38    0.349  -1.75          0.252 7.31e-04  5.20e-08    0.235
25    40    0.353  -1.72          0.242 5.67e-04  1.53e-08    0.200

cex1=0.9
plot(sft$fitIndices[,1], -sign(sft$fitIndices[,3])*sft$fitIndices[,2],xlab="Soft Threshold (power)",ylab="Scale Free Topology Model Fit, signed R^2",type="n", main = paste("Scale independence"))
text(sft$fitIndices[,1], -sign(sft$fitIndices[,3])*sft$fitIndices[,2],labels=powers ,cex=cex1,col="red")
abline(h=0.80,col="red")

#Mean Connectivity Plot
plot(sft$fitIndices[,1], sft$fitIndices[,5],xlab="Soft Threshold (power)",ylab="Mean Connectivity", type="n",main = paste("Mean connectivity"))
text(sft$fitIndices[,1], sft$fitIndices[,5], labels=powers, cex=cex1,col="red")

Generate TOM

softPower = 3
SubGeneNames<-colnames(datExpr)
adj= adjacency(datExpr, type = "signed", power = softPower)
diag(adj)<-0
TOM=TOMsimilarityFromExpr(datExpr, networkType = "signed", TOMType = "signed", power = softPower, maxPOutliers = 0.05)

TOM calculation: adjacency..
..will use 79 parallel threads.
 Fraction of slow calculations: 0.000000
..connectivity..
..matrix multiplication (system BLAS)..
..normalization..
..done.

colnames(TOM) = rownames(TOM) = SubGeneNames
dissTOM=1-TOM
geneTree = hclust(as.dist(dissTOM),method="average") #use complete for method rather than average (gives better results)
plot(geneTree,xlab="",sub="",cex=.5,main="Gene clustering",hang=.001)

Identify Modules

minModuleSize = 15
x = 4 
dynamicMods = cutreeDynamic(dendro = geneTree, distM = as.matrix(dissTOM), 
                            method="hybrid", pamStage = F, deepSplit = x, 
                            minClusterSize = minModuleSize)

 ..cutHeight not given, setting it to 0.874  ===>  99% of the (truncated) height range in dendro.
 ..done.

dynamicColors = labels2colors(dynamicMods) #label each module with a unique color
plotDendroAndColors(geneTree, dynamicColors, "Dynamic Tree Cut",
                    dendroLabels = FALSE, hang = 0.03, addGuide = TRUE, guideHang = 0.05, 
                    main = "Gene dendrogram and module colors") #plot the modules with colors

Calculate Eigengenes and Merge Close Modules

MEs = moduleEigengenes(datExpr, dynamicColors)$eigengenes #this matrix gives correlations between cells and module eigengenes (a high value indicates that the cell is highly correlated with the genes in that module)
ME1<-MEs
row.names(ME1)<-row.names(datExpr)
# Calculate dissimilarity of module eigengenes
MEDiss = 1-cor(MEs);
# Cluster module eigengenes
METree = hclust(as.dist(MEDiss), method = "average");
# Plot the result
plot(METree, main = "Clustering of module eigengenes",xlab = "", sub = "")
MEDissThres = 0.2
# Plot the cut line into the dendrogram
abline(h=MEDissThres, col = "red")

The merged module colors

merge = mergeCloseModules(datExpr, dynamicColors, cutHeight = MEDissThres, verbose = 3)

 mergeCloseModules: Merging modules whose distance is less than 0.2
   multiSetMEs: Calculating module MEs.
     Working on set 1 ...
     moduleEigengenes: Calculating 10 module eigengenes in given set.
   multiSetMEs: Calculating module MEs.
     Working on set 1 ...
     moduleEigengenes: Calculating 9 module eigengenes in given set.
   Calculating new MEs...
   multiSetMEs: Calculating module MEs.
     Working on set 1 ...
     moduleEigengenes: Calculating 9 module eigengenes in given set.

mergedColors = merge$colors
mergedMEs = merge$newMEs
moduleColors = mergedColors
MEs = mergedMEs
modulekME = signedKME(datExpr,MEs)

Plot merged modules

plotDendroAndColors(geneTree, cbind(dynamicColors, mergedColors),
c("Dynamic Tree Cut", "Merged dynamic"),
dendroLabels = FALSE, hang = 0.03,
addGuide = TRUE, guideHang = 0.05)

# Rename to moduleColors
moduleColors = mergedColors
# Construct numerical labels corresponding to the colors
# colorOrder = c("grey", standardColors(50));
# moduleLabels = match(moduleColors, colorOrder)-1
MEs = mergedMEs
modulekME = signedKME(datExpr,MEs)

#type gene name, prints out gene names also in that module
modules<-MEs
c_modules<-data.frame(moduleColors)
row.names(c_modules)<-colnames(datExpr) #assign gene names as row names
module.list.set1<-substring(colnames(modules),3) #removes ME from start of module names
index.set1<-0
Network=list() #create lists of genes for each module
for (i in 1:length(module.list.set1)){index.set1<-which(c_modules==module.list.set1[i])
Network[[i]]<-row.names(c_modules)[index.set1]}
names(Network)<-module.list.set1
lookup<-function(gene,network){return(network[names(network)[grep(gene,network)]])} #load function

Filter metadata table and correlate with eigengenes

nGenes = ncol(datExpr)
nSamples = nrow(datExpr)
MEs = orderMEs(MEs)
ventric$group<-paste0(ventric$trt,"_",ventric$day)
var<-model.matrix(~0+ventric$group)
#colnames(var)<-c("DV","FGF1","FGF19", "V")
moduleTraitCor <- cor(MEs, var, use="p")
moduleTraitPvalue = corPvalueStudent(moduleTraitCor, nSamples)
cor<-melt(moduleTraitCor)
ggplot(cor, aes(Var2, Var1)) + geom_tile(aes(fill = value), 
     colour = "white") + scale_fill_gradient2(midpoint = 0, low = "blue", mid = "white",
                            high = "red", space = "Lab", name="Correlation \nStrength") + 
  theme(axis.text.x = element_text(angle = 45, hjust = 1)) + xlab("Treatment") + ylab(NULL)

Get hubgenes in order

hubgenes<-lapply(seq_len(length(Network)), function(x) {
  dat<-modulekME[Network[[x]],]
  dat<-dat[order(-dat[paste0("kME",names(Network)[x])]),]
  gene<-rownames(dat)
  return(gene)
})
names(hubgenes)<-names(Network)
d <- unlist(hubgenes)
d <- data.frame(gene = d, 
           vec = names(d))
write_csv(d, path=here("output/glia/wgcna/allglia_wgcna_genemodules.csv"))

Build linear models for differential expression

MEs %>% select(-MEgrey) -> MEs
data<-data.frame(MEs, day=ventric$day, trt=ventric$trt, 
                 sample=as.factor(ventric$sample), group=ventric$group, 
                 batch=ventric$batch, celltype=Idents(ventric), 
                 groupall=paste0(Idents(ventric), ventric$group))

mod<-lapply(colnames(MEs), function(me) {
  mod<-lmer(MEs[[me]] ~ group*celltype + (1|batch) + (1|sample), data=data)
  pairwise<-emmeans(mod, pairwise ~ group|celltype)
  plot<-data.frame(plot(pairwise, plotIt=F)$data)
  sig<-as.data.frame(pairwise$contrasts)
  sig%>%separate(contrast, c("start", "end"), sep = " - ") -> sig
  yvals<-unlist(lapply(unique(sig$celltype), function(x) {
    x<-as.character(x)
    y<-data[data$celltype==x,]
    z<-max(as.numeric(y[[me]]))
    names(z)<-x
    return(z)
  }))
  sig$yvals<-yvals[match(sig$celltype, names(yvals))]
  sig$yvals[duplicated(sig$yvals)]<-sig$yvals[duplicated(sig$yvals)]+.004
  sig$yvals[duplicated(sig$yvals)]<-sig$yvals[duplicated(sig$yvals)]+.004
  sig$yvals[duplicated(sig$yvals)]<-sig$yvals[duplicated(sig$yvals)]+.004
  return(sig)
})

Warning in checkConv(attr(opt, "derivs"), opt$par, ctrl =
control$checkConv, : Model failed to converge with max|grad| = 0.00359644
(tol = 0.002, component 1)

names(mod) <- colnames(MEs)
sig <- bind_rows(mod, .id="id")
sig$symbol <- sig$p.value
sig$symbol[findInterval(sig$symbol, c(0.1,2)) == 1L] <-NA
sig$symbol[findInterval(sig$symbol, c(0.01,0.1)) == 1L] <- "*"
sig$symbol[findInterval(sig$symbol, c(0.001,0.01)) == 1L] <- "**"

Warning in findInterval(sig$symbol, c(0.001, 0.01)): NAs introduced by
coercion

sig$symbol[findInterval(sig$symbol, c(1e-200,0.001)) == 1L] <- "***" 

Warning in findInterval(sig$symbol, c(1e-200, 0.001)): NAs introduced by
coercion

data <- melt(data, id.vars = c("day","trt","sample","group","batch","celltype","groupall"))

lapply(unique(data$variable), function(x) {
  tryCatch({
  print(ggplot(data=data[data$variable==x,], aes(x=group, y=as.numeric(value))) + 
  geom_quasirandom(aes(fill=sample), shape=21, size=2, alpha=.75) + 
  scale_fill_manual(values=pal_jco()(10)) +  ylab(NULL) + xlab(NULL) + 
  theme_pubr() + theme(axis.text.x = element_text(angle=45, hjust=1, face="bold"), plot.title = element_text(hjust=0.5)) + 
  scale_y_continuous(aes(name="",limits=c(min(value)-.02,max(value))+.02)) + facet_wrap(.~celltype) +
  labs(y=NULL, x=NULL) + ggtitle(x)) },
  error = function(err) {
    print(err)
  }
  )
})

[[1]]

[[2]]

[[3]]

[[4]]

[[5]]

[[6]]

[[7]]

[[8]]

# moddat <- bind_rows(mod, .id="id")
write_csv(sig, path=here("output/glia/wgcna/allglia_wgcna_linearmodel_testing.csv"))

sig %>%
  unite(start, end, col = "comparison", remove = F) %>%
  filter(comparison == "FGF_Day-1_PF_Day-1" | comparison == "FGF_Day-5_PF_Day-5") %>%
  unite(estimate, p.value, sep = ",", col = "value") %>%
  dcast(id + celltype ~ start, value.var = "value") %>%
  separate(`FGF_Day-1`, into = c("estimate_1", "p.value_1"), sep = ",") %>%
  separate(`FGF_Day-5`, into = c("estimate_5", "p.value_5"), sep = ",") %>%
  mutate(id = gsub(id, pattern = "ME", replacement = "")) %>%
  mutate(col = if_else(as.numeric(p.value_1) < 0.1 | as.numeric(p.value_5) < 0.1, true = id, false = "grey")) %>%
  mutate(sig = if_else(as.numeric(p.value_1) < 0.1 & as.numeric(p.value_5) < 0.1, true = "red",
    false = if_else(as.numeric(p.value_1) < 0.1, true = "blue",
      false = if_else(as.numeric(p.value_5) < 0.1, true = "black", false = "")
    )
  )) %>%
  mutate(label = if_else(sig == "blue", true = "DE at Day 1", false = "")) -> plot

cols <- unique(plot$col)
names(cols) <- cols

ggplot(plot, aes(x = as.numeric(estimate_1), y = as.numeric(estimate_5))) +
  geom_point(data = filter(plot, sig != ""), aes(shape = celltype), size = 4) +
  scale_shape(name="Cell-Type") +
  geom_mark_ellipse(data = plot %>% filter(sig == "blue", p.value_1 != 0.0230780280327769), color = "black", linetype = "dashed") +
  geom_point(size = 4, aes(color = col, shape = celltype)) + coord_flip() + geom_hline(yintercept = 0) +
  scale_color_manual(values = cols, name = "Modules") +
  geom_vline(xintercept = 0) +
  annotate(
    geom = "curve", x = 0.045, y = 0.02, xend = 0.03, yend = 0.005,
    curvature = .3, arrow = arrow(length = unit(2, "mm"))
  ) + annotate(geom = "label", x = 0.045, y = 0.022, label = "DE only at Day 1", hjust = "left") +
  annotate(
    geom = "curve", x = -0.01, y = 0.03, xend = -0.025, yend = 0.035,
    curvature = .3, arrow = arrow(length = unit(2, "mm"))
  ) + annotate(geom = "label", x = -0.005, y = 0.0275, label = "DE at both", hjust = "left") +
  annotate(
    geom = "curve", x = 0.01, y = 0.015, xend = -0.011, yend = 0.02,
    curvature = .3, arrow = arrow(length = unit(2, "mm"))
  ) + annotate(geom = "label", x = 0.015, y = 0.01, label = "DE only at Day 5", hjust = "left") + theme_bw() + xlab("Day 1 Estimate") +
  ylab("Day 5 Estimate") + labs_pubr()

ggsave(here("output/wgcna_res.png"), w=8, h=5)

Find GO term enrichment of modules

names(mod)<-colnames(MEs)[grepl("^ME", colnames(MEs))]
go_col <- unique(plot$col)[-grep("grey", unique(plot$col))]
goterms<-lapply(go_col, function(x) { 
  x<-gprofiler(hubgenes[[x]], ordered_query = T, 
               organism = "mmusculus", significant = T, custom_bg = colnames(datExpr),
                           src_filter = c("GO:BP","REAC","KEGG"), hier_filtering = "strong",
                           min_isect_size = 3, 
                           sort_by_structure = T,exclude_iea = T, 
                           min_set_size = 10, max_set_size = 300,correction_method = "fdr")
  x<-x[order(x$p.value),]
  return(x)
})

names(goterms) <- go_col
goterms %>% bind_rows(.id="id") %>%
  mutate(padj=p.adjust(p.value, "fdr")) -> godat

write_csv(godat, path=here("output/glia/wgcna/allglia_wgcna_goterms.csv"))

godat %>% group_by(id) %>% filter(id %in% c("black","green")) %>% arrange(p.value) %>% slice(1:5) %>% 
  select(p.value, padj, term.name, domain, id) %>% arrange(id) %>%
  ggplot(aes(x=str_wrap(term.name,20), y=-log10(padj), fill=domain)) + geom_col() +
  scale_fill_npg() +
  facet_wrap(.~id, scales="free_y", ncol=1) + theme_pubr() + 
  theme(text = element_text(size=7), 
        axis.text.x = element_text(angle=45, hjust=1)) + coord_flip() +
  xlab("GO Term") + geom_hline(yintercept = -log10(0.05), linetype="dashed", size=1) + 
  labs_pubr()

ggsave(filename = here("output/glia/wgcna/allglia_goterm.png"), h=6, w=8)

Session information

sessionInfo()

R version 3.5.3 (2019-03-11)
Platform: x86_64-pc-linux-gnu (64-bit)
Running under: Storage

Matrix products: default
BLAS/LAPACK: /usr/lib64/libopenblas-r0.3.3.so

locale:
 [1] LC_CTYPE=en_DK.UTF-8       LC_NUMERIC=C              
 [3] LC_TIME=en_DK.UTF-8        LC_COLLATE=en_DK.UTF-8    
 [5] LC_MONETARY=en_DK.UTF-8    LC_MESSAGES=en_DK.UTF-8   
 [7] LC_PAPER=en_DK.UTF-8       LC_NAME=C                 
 [9] LC_ADDRESS=C               LC_TELEPHONE=C            
[11] LC_MEASUREMENT=en_DK.UTF-8 LC_IDENTIFICATION=C       

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base     

other attached packages:
 [1] ggforce_0.3.0.9000    here_0.1              ggpubr_0.2.1         
 [4] magrittr_1.5          gProfileR_0.6.7       igraph_1.2.4.1       
 [7] reshape2_1.4.3        forcats_0.4.0         stringr_1.4.0        
[10] dplyr_0.8.3           purrr_0.3.2           readr_1.3.1.9000     
[13] tidyr_0.8.3           tibble_2.1.3          tidyverse_1.2.1      
[16] genefilter_1.64.0     ggbeeswarm_0.6.0      ggplot2_3.2.1        
[19] lme4_1.1-21           Matrix_1.2-17         emmeans_1.3.5.1      
[22] ggsci_2.9             parallelDist_0.2.4    cluster_2.1.0        
[25] WGCNA_1.68            fastcluster_1.1.25    dynamicTreeCut_1.63-1
[28] Seurat_3.0.3.9036    

loaded via a namespace (and not attached):
  [1] reticulate_1.13       R.utils_2.9.0         tidyselect_0.2.5     
  [4] robust_0.4-18.1       RSQLite_2.1.1         AnnotationDbi_1.44.0 
  [7] htmlwidgets_1.3       grid_3.5.3            Rtsne_0.15           
 [10] munsell_0.5.0         codetools_0.2-16      ica_1.0-2            
 [13] preprocessCore_1.44.0 future_1.14.0         withr_2.1.2          
 [16] colorspace_1.4-1      Biobase_2.42.0        highr_0.8            
 [19] knitr_1.23            rstudioapi_0.10       stats4_3.5.3         
 [22] ROCR_1.0-7            robustbase_0.93-5     ggsignif_0.5.0       
 [25] gbRd_0.4-11           listenv_0.7.0         labeling_0.3         
 [28] Rdpack_0.11-0         git2r_0.25.2          polyclip_1.10-0      
 [31] farver_1.1.0          bit64_0.9-7           rprojroot_1.3-2      
 [34] vctrs_0.2.0           coda_0.19-3           generics_0.0.2       
 [37] TH.data_1.0-10        xfun_0.8              R6_2.4.0             
 [40] doParallel_1.0.14     rsvd_1.0.2            bitops_1.0-6         
 [43] assertthat_0.2.1      SDMTools_1.1-221.1    scales_1.0.0         
 [46] multcomp_1.4-10       nnet_7.3-12           beeswarm_0.2.3       
 [49] gtable_0.3.0          npsurv_0.4-0          globals_0.12.4       
 [52] sandwich_2.5-1        workflowr_1.4.0       rlang_0.4.0          
 [55] zeallot_0.1.0         splines_3.5.3         lazyeval_0.2.2       
 [58] acepack_1.4.1         impute_1.56.0         broom_0.5.2          
 [61] checkmate_1.9.4       modelr_0.1.4          yaml_2.2.0           
 [64] backports_1.1.4       Hmisc_4.2-0           tools_3.5.3          
 [67] gplots_3.0.1.1        RColorBrewer_1.1-2    BiocGenerics_0.28.0  
 [70] ggridges_0.5.1        Rcpp_1.0.2            plyr_1.8.4           
 [73] base64enc_0.1-3       RCurl_1.95-4.12       rpart_4.1-15         
 [76] pbapply_1.4-1         cowplot_1.0.0         S4Vectors_0.20.1     
 [79] zoo_1.8-6             haven_2.1.0           ggrepel_0.8.1        
 [82] fs_1.3.1              data.table_1.12.2     lmtest_0.9-37        
 [85] RANN_2.6.1            mvtnorm_1.0-11        fitdistrplus_1.0-14  
 [88] matrixStats_0.54.0    hms_0.5.0             lsei_1.2-0           
 [91] evaluate_0.14         xtable_1.8-4          pbkrtest_0.4-7       
 [94] XML_3.98-1.20         readxl_1.3.1          IRanges_2.16.0       
 [97] gridExtra_2.3         compiler_3.5.3        KernSmooth_2.23-15   
[100] crayon_1.3.4          minqa_1.2.4           R.oo_1.22.0          
[103] htmltools_0.3.6       pcaPP_1.9-73          Formula_1.2-3        
[106] rrcov_1.4-7           RcppParallel_4.4.3    lubridate_1.7.4      
[109] DBI_1.0.0             tweenr_1.0.1          MASS_7.3-51.4        
[112] boot_1.3-22           cli_1.1.0             R.methodsS3_1.7.1    
[115] gdata_2.18.0          parallel_3.5.3        metap_1.1            
[118] pkgconfig_2.0.2       fit.models_0.5-14     foreign_0.8-71       
[121] plotly_4.9.0          xml2_1.2.0            foreach_1.4.4        
[124] annotate_1.60.1       vipor_0.4.5           estimability_1.3     
[127] rvest_0.3.4           bibtex_0.4.2          digest_0.6.20        
[130] sctransform_0.2.0     RcppAnnoy_0.0.12      tsne_0.1-3           
[133] cellranger_1.1.0      rmarkdown_1.13        leiden_0.3.1         
[136] htmlTable_1.13.1      uwot_0.1.3            gtools_3.8.1         
[139] nloptr_1.2.1          nlme_3.1-140          jsonlite_1.6         
[142] viridisLite_0.3.0     pillar_1.4.2          lattice_0.20-38      
[145] httr_1.4.1            DEoptimR_1.0-8        survival_2.44-1.1    
[148] GO.db_3.7.0           glue_1.3.1            png_0.1-7            
[151] iterators_1.0.10      bit_1.1-14            stringi_1.4.3        
[154] blob_1.1.1            latticeExtra_0.6-28   caTools_1.17.1.2     
[157] memoise_1.1.0         irlba_2.3.3           future.apply_1.3.0   
[160] ape_5.3