the integration results from liger object and seurat object are different

[shenscore]

I followed the code from the usage page.

I found the UMAP results are different:

start from liger object : https://welch-lab.github.io/liger/articles/Integrating_multi_scRNA_data.html
start from seurat object : https://welch-lab.github.io/liger/articles/liger_with_seurat.html#move-on-to-seurat-downstream-analysis-after-liger

[mvfki]

Because many functions and parameters used during the whole procedure are different.

[shenscore]

Because many functions and parameters used during the whole procedure are different.

Yeah. So I run these codes instead:

For liger object:

sni.liger <- sni.liger %>%
  normalize() %>%
  selectGenes() %>%
  scaleNotCenter()

sni.liger <- runIntegration(sni.liger, k = 30)
sni.liger <- alignFactors(sni.liger, method = "centroidAlign")
sni.liger <- runCluster(sni.liger, resolution = 0.3)
sni.liger <- runUMAP(sni.liger, minDist = 0.3)

For seurat object:

sni.sr.data[["RNA"]] <- split(sni.sr.data[["RNA"]],f = sni.sr.data$Stage)
sni.sr.data <- sni.sr.data %>%
  normalize() %>%
  selectGenes() %>%
  scaleNotCenter()

sni.sr.data <- runINMF(sni.sr.data, k = 30) %>% centroidAlign()
sni.sr.data <- FindClusters(sni.sr.data, algorithm = 4,cluster.name = "leiden_cluster")
sni.sr.data <- RunUMAP(sni.sr.data, reduction = "inmfNorm",dims=1:30, reduction.name = "umap.liger")

I think I used the same functions. I don't know which parameter causes the different results.