本来呢，单细胞UMAP或者TSNE降维图正常展示就好了，最多修改颜色坐标就可以，但是奈何总是有文章挑战别人的眼睛，非要秀一波非常规操作，那我们也只能把他们打下神坛。
事情的起因是看了这篇文章中的UMAP图，有啥特点呢？每个点都是有边框的，这不就是ggplot中选择shape=21的形状吗？很显然，Dimplot是实现不了的。UMAP降维图本质也是散点图，只需要将作图数据导出，ggplot2就可以实现任何你想要的修饰了。
image.png
首先我们设置下颜色，并将作图的数据导出，导出的数据包含UMAP两个坐标的数据。
allcolour=c("#20B2AA","#FFA500","#9370DB","#98FB98","#1E90FF","#7CFC00","#FFFF00",
"#808000","#FF00FF","#FA8072","#7B68EE","#9400D3","#800080","#A0522D","#D2B48C","#D2691E","#87CEEB","#40E0D0","#5F9EA0",
"#FF1493","#0000CD","#008B8B","#FFE4B5","#8A2BE2","#228B22","#E9967A","#4682B4","#32CD32","#F0E68C","#FFFFE0","#EE82EE",
"#FF6347","#6A5ACD","#9932CC","#8B008B","#8B4513","#DEB887")
umap = NCsce@reductions$umap@cell.embeddings %>%
as.data.frame() %>%
cbind(cell_type = NCsce@meta.data$celltype)
然后就可以绘图了，很简单：
ggplot(umap, aes(x= UMAP_1 , y = UMAP_2 ,fill = cell_type))+
geom_point(size = 3.5,colour="grey35",shape=21) +
scale_fill_manual(values = allcolour)+
theme_classic()
image.png
可以像有些文章中一样，干脆不显示坐标轴，就是单纯的图。
ggplot(umap, aes(x= UMAP_1 , y = UMAP_2 ,fill = cell_type))+
geom_point(size = 3,colour="grey35",shape=21) +
scale_fill_manual(values = allcolour)+
theme_bw()+
theme(panel.grid =element_blank(),
panel.border = element_blank(),
axis.text.y = element_blank(),
axis.ticks.y = element_blank(),
axis.ticks.x = element_blank(),
axis.text.x = element_blank())+
labs(x='', y='')
image.png
好了，这里也就是仅仅说明有这么一种情况存在，想要继续修饰可以发挥你们的聪明才智，争取有一个惊艳众人的可视化结果。
