复现Nature图表：单细胞UMAP图指示celltype和ma

看到了一篇Nature的UMAP图，展示的是单细胞，但是特点是标记了celltype的同时指示了marker基因。显然这个处理很有特色，我们可以用代码指示，但是很难一次性达到这个效果，最后借助AI可能会更加方便，一次性完成这个事情，关于AI的操作，我们是直接整了个视频，希望对你有用！

（reference：IL-1β+ macrophages fuel pathogenic inflammation in pancreatic cancer)

普通的UMAP是这样的：

setwd('D:\\KS项目\\公众号文章\\复现nature单细胞UMAP降维图修饰')
library(Seurat)
library(ggplot2)
library(ggrastr)
library(tidydr)
library(dplyr)
library(ggrepel)

# load("C:/Users/tq199/Downloads/adj_scRNA.RData")
adj_scRNA <- subset(adj_scRNA, celltype=="Other", invert=T)
DimPlot(adj_scRNA, label = T)

image.png

之后我们提取坐标，用ggplot修饰一下：

df <- adj_scRNA@reductions$umap@cell.embeddings%>% 
  as.data.frame() %>%
  cbind(cell_type = adj_scRNA@meta.data$celltype)

label <- df %>%group_by(cell_type) %>%
  summarise(UMAP_1 = median(UMAP_1),
            UMAP_2 = median(UMAP_2))%>%
  as.data.frame()
rownames(label) <- label$cell_type

label$number <- seq(1:9)

cols= c('#7F3C8D' ,'#11A579', '#3969AC',
        '#E73F74', '#80BA5A', '#E68310',
        '#008695', '#CF1C90', '#f97b72')

#ggplot作图
p = ggplot()+
  geom_point_rast(data=df, aes(x= UMAP_1 , y = UMAP_2 ,color = cell_type),size = 1,shape=16) +
  scale_color_manual(values = alpha(cols,0.3))+ #设置下透明度
  theme_classic()+
  theme(panel.grid.major = element_blank(),
        panel.grid.minor = element_blank(),
        axis.ticks = element_blank(),
        axis.line = element_blank(),
        axis.title = element_blank(),
        axis.text = element_blank(),
        legend.position = 'none')+
  geom_point(data = label, aes(x= UMAP_1 , y = UMAP_2), size=6, color='white', alpha=0.9)+
  geom_point(data = label, aes(x= UMAP_1 , y = UMAP_2), size=6, color='black', shape=21)+
  geom_text(data = label,
            mapping = aes(x= UMAP_1 , y = UMAP_2, label = number),
            color='black')

image.png

然后标记基因即可：

marker_gene <- FindAllMarkers(adj_scRNA, only.pos = T,
                              logfc.threshold = 0.8, min.pct = 0.8)

#每种celltype挑选5个展示
marker_genes <- marker_gene %>% 
  group_by(cluster) %>% 
  top_n(n = 5, wt = avg_log2FC) 

write.csv(marker_genes, file = 'marker_genes.csv')

p + geom_label_repel(
  data=label[which(label$UMAP_1>0 &label$UMAP_2>0),], 
  aes(x= UMAP_1, y = UMAP_2,label=labels), 
  size=3,
  nudge_x = 4,
  box.padding = 0.5,
  nudge_y = 8,
  segment.curvature = -0.1,
  segment.ncp = 3,
  segment.angle = 20,
  direction = "y", hjust = "left")+
  geom_label_repel(
    data=label[which(label$UMAP_1>0 &label$UMAP_2<0),], 
    aes(x= UMAP_1, y = UMAP_2,label=labels), 
    size=3,
    nudge_x = 4,
    box.padding = 0.5,
    nudge_y = -8,
    segment.curvature = -0.1,
    segment.ncp = 3,
    segment.angle = 20,
    direction = "y", hjust = "left")+
  geom_label_repel(
    data=label[which(label$UMAP_1<0 &label$UMAP_2>3),], 
    aes(x= UMAP_1, y = UMAP_2,label=labels), 
    size=3,
    nudge_x = -6,
    box.padding = 0.5,
    nudge_y = 15,
    segment.curvature = -0.1,
    segment.ncp = 3,
    segment.angle = 20,
    direction = "y", hjust = "left")+
  geom_label_repel(
    data=label[which(label$UMAP_1<0 &label$UMAP_2<3),], 
    aes(x= UMAP_1, y = UMAP_2,label=labels), 
    size=3,
    nudge_x = -3,
    box.padding = 0.5,
    nudge_y = -12,
    segment.curvature = -0.1,
    segment.ncp = 3,
    segment.angle = 20,
    direction = "y", hjust = "left")

image.png

很显然，代码不能一次性，至少我没有方法能够一次性达到这种效果。所以遇到这种，直接上AI或者PS，省时省力！

餐卡：
https://mp.weixin.qq.com/s?__biz=Mzg5OTYzMzY5Ng==&mid=2247489017&idx=1&sn=ea3bfd43447edfb372636a3412c712fe&chksm=c05112b6f7269ba05d3d295d389d0ae1e368994e7169d8cac601b7e210140bd9ba51102ae682&token=512044395&lang=zh_CN#rd