1. 模块划分（聚类）

Signed or unsigned: which network type is preferable?

net <- blockwiseModules(
  datExpr, # 输入数据
  corType = "pearson", # pearson | bicor
  power = sft$powerEstimate, # 前面得到的软阈值
  networkType = "unsigned", # unsigned | signed | signed hybrid
  TOMType = "unsigned", # none | unsigned | signed
  saveTOMs = TRUE,
  saveTOMFileBase = "blockwiseTOM",
  # deepSplit = 2, # 0|1|2|3|4, 值越大得到的模块就越多越小
  minModuleSize = 30, # 模块中最少的基因数
  mergeCutHeight = 0.25, # 对距离小于 mergeCutHeight （1-cor）的模块进行合并
  numericLabels = FALSE, # 以数字命名模块
  nThreads = 0, # 0 则使用所有可用线程
  maxBlockSize = nGenes # 需要大于基因的数目
)
net %>% names()
#> [1] "colors"         "unmergedColors" "MEs"            "goodSamples"
#> [5] "goodGenes"      "dendrograms"    "TOMFiles"       "blockGenes"
#> [9] "blocks"         "MEsOK"
# net$MEs: 模块特征向量
wgcna_result <- data.frame(gene_id = names(net$colors), module = net$colors)
moduleColors = (net$colors)
pdf(file = "WGCNA_result.pdf")
plotDendroAndColors(dendro = net$dendrograms[[1]], 
                    colors = net$colors,
                    groupLabels = "Module colors",
                    dendroLabels = FALSE,
                    addGuide = TRUE)
dev.off()

2. 相似模块合并

3. 模块特征向量

# my_order: 样本组别顺序
tgc <- net$MEs %>%
  rownames_to_column("Sample") %>%
  pivot_longer(-Sample, names_to = "Module", values_to = "Vleu") %>%
  mutate(Type = str_replace(Sample, "\\..*", "")) %>% # 得到样本组别信息，根据自己情况
  arrange(Module) %>%
  summarySE(measurevar="Vleu", groupvars=c("Module","Type")) %>%
  mutate(Type = factor(Type, levels = my_order)) %>%
  arrange(Module, Type)
pdf(file='gene_expression_pattern_in_all_modules_1.pdf', w=12,h=14)
ggplot(tgc, aes(Type, Vleu, group = Module)) +
  geom_errorbar(aes(ymin = Vleu - se, ymax = Vleu + se), width = .1, color = "#99CCCC") +
  geom_line(color = "#99CCCC", size = 0.5) +
  geom_point(color = "#99CCCC") +
  facet_wrap(~ Module, ncol = 4) +
  theme_bw()
dev.off()