寻找差异基因到底使用哪个？FindMarkers、FindallMarkers 、 findconservedmarkers 、还是 PrepSCTFindMarkers

Q：得到的差异基因和保守基因，用哪个作为 marker 基因好？ / FindallMarkers 和findconservedmarkers 结果上有什么区别吗？

A：如果只有一个样本，就用 cluster 中的差异表达基因做 marker 基因。如果做了数据整合，比如处理组和对照组，就要先用 FindConservedMarkers 函数找到保守的差异表达基因，用它们来作为 marker 基因对 cluster 做注释，注释完以后，再用 FindMarkers 函数比较这个 cluster 中的处理组细胞和对照组细胞之间的差异表达基因。

需要解答：那要是不止两组，有好几组，用哪个函数呢？解决方案___

转自：如何使用 Seurat 分析单细胞测序数据（ Q&A - 简书 (jianshu.com)

FindConservedMarkers实例
Seurat之整合分析（1） - 简书 (jianshu.com)
2020-05-18 seurat 包 Stimulated vs Control PBMCs - 简书 (jianshu.com)
单细胞测序数据整合分析示例 - 简书 (jianshu.com)
10x Genomics PBMC（七）：整合数据后的聚类分析 - 简书 (jianshu.com)

2023.05.11

昨天找到了一篇帖子：【细胞通讯】PlantPhoneDB（2），是植物的样本，实验组和对照组。用SCT标准化，FindIntegrationAnchors做了去批次。作者在做完FindClusters之后，关键的两步是DefaultAssay(objs) <- 'SCT' 和 objs<- PrepSCTFindMarkers(objs,assay = "SCT", verbose = TRUE) 贴一下作者的代码：先按照这个跑一下看看

library(Seurat)

library(tidyverse)

library(ggplot2)

library(ggsci)

library(ggpubr)

library(pheatmap)

library(RColorBrewer)

library(patchwork)

library(lsa)

library(viridis)

library(hrbrthemes)

library(circlize)

library(chorddiag)

library(ggplotify)

library(data.table)

library(parmigene)

library(readxl)

library(infotheo)

library(igraph)

library(muxViz)

library(rgl)

library(tidyverse)

library(dplyr)

//参数选择和作者paper中用的参数一样

pbmc <- readRDS("pbmc3k_final.rds")

pbmc@meta.data$labels <- Idents(pbmc)

control.data <- Read10X(data.dir = "control/filtered_feature_bc_matrix/")

control<- CreateSeuratObject(counts = control.data, project = "control", min.cells = 3, min.features = 200)

control <- subset(control, subset = nFeature_RNA > 200 & nCount_RNA > 1000)

control <- SCTransform(control, verbose = FALSE)

heat.data <- Read10X(data.dir = "heat/filtered_feature_bc_matrix/")

heat<- CreateSeuratObject(counts = heat.data, project = "heat", min.cells = 3, min.features = 200)

heat <- subset(heat, subset = nFeature_RNA > 200 & nCount_RNA > 1000)

heat <- SCTransform(heat, verbose = FALSE)

datasets <- c(control,heat)

features <- SelectIntegrationFeatures(object.list = datasets, nfeatures = 8000)

datasets <- PrepSCTIntegration(object.list = datasets, anchor.features = features, verbose = TRUE)

datasets <- lapply(X = datasets, FUN = RunPCA, verbose = FALSE, features = features)

anchors <- FindIntegrationAnchors(object.list = datasets, normalization.method = "SCT",anchor.features = features, verbose = TRUE, reference=1,reduction = "cca")

objs <- IntegrateData(anchorset = anchors, normalization.method = "SCT", verbose = TRUE)

objs <- RunPCA(objs, verbose = FALSE, approx = FALSE, npcs = 50)

objs <- RunUMAP(objs, reduction = "pca", dims = 1:50, umap.method = "umap-learn", metric = "correlation")

objs <- RunTSNE(objs, reduction = "pca",dims = 1:50,tsne.method = "Rtsne")

objs <- FindNeighbors(objs, reduction = "pca",dims = 1:50)

objs <- FindClusters(objs, resolution = 0.5, algorithm = 2)

DefaultAssay(objs) <- 'SCT'

objs<- PrepSCTFindMarkers(objs,assay = "SCT", verbose = TRUE)

DEG <- FindAllMarkers(objs,

                      logfc.threshold=0.25,

                      min.diff.pct = 0.25,

                      max.cells.per.ident = 10000,

                      only.pos=T)

mark_gene <- DEG %>% mutate(avg_logFC=avg_log2FC) %>% filter(p_val_adj<0.05)

signature <- readxl::read_excel('../ath_doi_202104.xlsx')

sig_gene  <- signature %>% as.data.frame() %>% filter(Tissue=="Root") %>% mutate(V1=`Cell Type`,V2=Cell_Marker) %>% unique(.) %>% select(V1,V2)

贴一下报错，搜了一下ggrepel: 7 unlabeled data points (too many overlaps). Consider increasing max.overlaps ，说不要紧

 DefaultAssay(integrated) <- 'SCT'
> 
> integrated<- PrepSCTFindMarkers(integrated,assay = "SCT", verbose = TRUE)
Found 6 SCT models. Recorrecting SCT counts using minimum median counts: 2575
> 
> DEG <- FindAllMarkers(integrated,
+                       
+                       logfc.threshold=0.25,
+                       
+                       min.diff.pct = 0.25,
+                       
+                       only.pos=T)
Calculating cluster 0
  |++++++++++++++++++++++++++++++++++++++++++++++++++| 100% elapsed=02m 41s
Calculating cluster 1
  |++++++++++++++++++++++++++++++++++++++++++++++++++| 100% elapsed=33s  
Calculating cluster 2
  |++++++++++++++++++++++++++++++++++++++++++++++++++| 100% elapsed=13s  
Calculating cluster 3
  |++++++++++++++++++++++++++++++++++++++++++++++++++| 100% elapsed=39s  
Calculating cluster 4
  |++++++++++++++++++++++++++++++++++++++++++++++++++| 100% elapsed=20s  
Calculating cluster 5
  |++++++++++++++++++++++++++++++++++++++++++++++++++| 100% elapsed=04s  
Calculating cluster 6
  |++++++++++++++++++++++++++++++++++++++++++++++++++| 100% elapsed=50s  
Calculating cluster 7
  |++++++++++++++++++++++++++++++++++++++++++++++++++| 100% elapsed=21s  
Calculating cluster 8
  |++++++++++++++++++++++++++++++++++++++++++++++++++| 100% elapsed=02m 24s
Calculating cluster 9
  |++++++++++++++++++++++++++++++++++++++++++++++++++| 100% elapsed=52s  
Calculating cluster 10
  |++++++++++++++++++++++++++++++++++++++++++++++++++| 100% elapsed=29s  
Calculating cluster 11
  |++++++++++++++++++++++++++++++++++++++++++++++++++| 100% elapsed=30s  
Calculating cluster 12
  |++++++++++++++++++++++++++++++++++++++++++++++++++| 100% elapsed=04m 21s
Calculating cluster 13
  |++++++++++++++++++++++++++++++++++++++++++++++++++| 100% elapsed=19s  
Calculating cluster 14
  |++++++++++++++++++++++++++++++++++++++++++++++++++| 100% elapsed=16s  
Calculating cluster 15
  |++++++++++++++++++++++++++++++++++++++++++++++++++| 100% elapsed=02m 02s
Calculating cluster 16
  |++++++++++++++++++++++++++++++++++++++++++++++++++| 100% elapsed=01m 47s
Calculating cluster 17
  |++++++++++++++++++++++++++++++++++++++++++++++++++| 100% elapsed=03m 22s
Calculating cluster 18
  |++++++++++++++++++++++++++++++++++++++++++++++++++| 100% elapsed=48s  
Calculating cluster 19
  |++++++++++++++++++++++++++++++++++++++++++++++++++| 100% elapsed=28s  
Calculating cluster 20
  |++++++++++++++++++++++++++++++++++++++++++++++++++| 100% elapsed=01m 53s
Calculating cluster 21
  |++++++++++++++++++++++++++++++++++++++++++++++++++| 100% elapsed=54s  
Calculating cluster 22
  |++++++++++++++++++++++++++++++++++++++++++++++++++| 100% elapsed=01m 11s
Calculating cluster 23
  |++++++++++++++++++++++++++++++++++++++++++++++++++| 100% elapsed=01m 05s
Warning messages:
1: ggrepel: 7 unlabeled data points (too many overlaps). Consider increasing max.overlaps 
2: In UseMethod("depth") :
  no applicable method for 'depth' applied to an object of class "NULL"
3: In UseMethod("depth") :
  no applicable method for 'depth' applied to an object of class "NULL"
4: In UseMethod("depth") :
  no applicable method for 'depth' applied to an object of class "NULL"
>