https://cloud.tencent.com/developer/article/1813350
1. Feature loading 结果解读
https://www.jianshu.com/p/bcb0b520e056
loading值越大，则此基因对此PC的贡献越大
2. 计算 PVE
PVE: Percent Variance Explained
每个
PC对总体变异的解释比例
object_used <- Human_REP1_RPL35_KD_25uM_3_200_filtered_feature_over_2500_HVF_1000_dim_5_with_cell_ctcle
(object_used[["pca"]]@stdev^2/sum(object_used[["pca"]]@stdev^2)) * 100
3. 报错及解决一：PCA duplicates 导致 tSNE 报错
报错：
Error in Rtsne.default(X = object, dims=dim.embed, pca=FALSE, ...) : Remove duplicates before running TSNE.
报错原因：
数据中出现
PCA结果完全相同的细胞(duplicate)
Feature越少，比如对HTO数据进行PCA然后再tSNE，duplicate出现的可能性越高
解决：
RunTSNE(..., check_duplicates=F): 强制软件在运行的时候忽略duplicates
网友评论