
popV 概述
Fig. 1: Framework of popV for automatic cell type annotation.

PopV预测得分区分高质量和低质量注释
Fig. 2: PopV prediction on LCA and TS lung as reference is accurate and interpretable.

PopV 在细胞组成存在显著差异的情况下提供了有用的标签转移
Fig. 3: PopV identifies thymocytes as query-specific cell types and yields highly interpretable consensus scores.

数据集
智者之表
肺细胞图谱
脑数据集
胸腺数据集
核测序和Drop-seq数据集
模型参数
预处理
[ol]- 1. retrain—it trains all methods from scratch and stores the classifier to reuse them on other datasets. This hugely benefits from a GPU to train the scVI and scANVI algorithms as well as the OnClass algorithm. - 2. inference—it uses pretrained methods to classify query and reference cells; computes a joint UMAP embedding of query and reference cells and by default uses all eight methods; and trains scVI and scANVI models for 20 epochs using scArches query embedding19. - 3. fast—it uses pretrained methods to classify only query cells; computes a UMAP embedding of query cells if enabled; skips Scanorama and BBKNN data integration as those recompute an embedding instead of projecting cells into an existing embedding; and trains scVI and scANVI models for 1 epoch using scArches query embedding.
BBKNN
全景图
scVI 是一种用于单细胞转录组数据的变分自编码器方法。
scANVI
射频
支持向量机
细胞类型学家
OnClass
和谐
塞尚标签转移
共识投票
评估指标
准确度指标
混淆矩阵
差异表达分析
精确-召回曲线
消融实验
与多数投票的SVM分类器进行比较
统计与可重复性
报告摘要