SLIDE 1
Batch effects
2 / 11
CellMixS Explore data integration and batch effects Almut Ltge - - PowerPoint PPT Presentation
CellMixS Explore data integration and batch effects Almut Ltge - - PowerPoint PPT Presentation
CellMixS Explore data integration and batch effects Almut Ltge DMLS - University of Zrich 2019-12-08 Batch effects 2 / 11 3 / 11 library (CellMixS) # Input data SingleCellExperiment object sce1 <-
SLIDE 2
SLIDE 3
3 / 11
SLIDE 4
library(CellMixS) # Input data SingleCellExperiment object sce1 <- readRDS("data/sim_batch_2.rds") sce2 <- readRDS("data/sim_batch_1.5.rds") # Get Cellspecific Mixing Score sce1 <- evalIntegration(metrics = "cms", sce1, group = "batch_id", k = 100) sce2 <- evalIntegration(metrics = "cms", sce2, group = "batch_id", k = 100)
4 / 11
SLIDE 5
Cellspecific Mixing Score
5 / 11
SLIDE 6
Cellspecific Mixing Score
# show groups visGroup(sce1, "batch_id") # show metric visMetric(sce1, "cms_smooth")
6 / 11
SLIDE 7
Interpretation
#get cms scores from sce cms_res <- data_frame("cms_sce1" = sim_batch2$cms, "cms_sce2" = sim_batch$cms) # show scores as histogram p <- visHist(cms_res, n_col = 2)
7 / 11
SLIDE 8
Random batches
#assign random batch label sce2$random_batch <- sample(sce2$batch_id, ncol(sce), replace = FALSE) # run cms sce2 <- evalIntegration("cms", sce = sce2, group = "random_batch", k = 100, res_name = "cms_random") # show scores as histogram cms_res$cms_random <- sce2$cms_random p <- visHist(cms_res, n_col = 3)
8 / 11
SLIDE 9
More options
9 / 11
SLIDE 10
More options
More functionalities
Adjust neighbourhoods (important for unbalanced batches) Group by celltype Compare Integration
More metrics
entrophy inverse Simpson Index mixingMetric (Seurat) Local density factor LocalStructure (Seurat)
10 / 11
SLIDE 11
Quantify (batch-) mixing Cellspecific Neighbourhood adjustment (Unbalanced batches!) Different metrics Quantify structual changes upon integration cellspecific Visualize mixing
Summary
11 / 11