27-11-2024 LOG: Annotating clusters (AinO) [manual, auto, label transfer, scVI]

[florekem]

Integration with HumanNewlyDiagnosed using scVI

[florekem]

But first, I'll try to do some more work with clustering.

SCT and different findneighbors/findclusters parameters:

1.

seu_singlet <- FindNeighbors(
  seu_singlet,
  reduction = "pca", dims = 1:20,
  k.param = 20 # (default)
)
seu_singlet <- FindClusters(
  seu_singlet,
  resolution = 0.5, verbose = FALSE,
  algorithm = 4 # leiden
)
seu_singlet <- RunUMAP(
  seu_singlet,
  reduction = "pca", dims = 1:20
)

2.

seu_singlet <- FindNeighbors(
  seu_singlet,
  reduction = "pca", dims = 1:20,
  **k.param = 40**
)
seu_singlet <- FindClusters(
  seu_singlet,
  resolution = 0.5, verbose = FALSE,
  algorithm = 4 # leiden
)
seu_singlet <- RunUMAP(
  seu_singlet,
  reduction = "pca", dims = 1:20
)

3.

seu_singlet <- FindNeighbors(
  seu_singlet,
  reduction = "pca", dims = 1:20,
  k.param = 40,
  annoy.metric = "manhattan"
)
seu_singlet <- FindClusters(
  seu_singlet,
  resolution = 0.5, verbose = FALSE,
  algorithm = 4 # leiden
)
seu_singlet <- RunUMAP(
  seu_singlet,
  reduction = "pca", dims = 1:20

4.

dims = 1:30
seu_singlet <- FindNeighbors(
  seu_singlet,
  reduction = "pca", dims = dims,
  k.param = 30,
  annoy.metric = "manhattan"
)
seu_singlet <- FindClusters(
  seu_singlet,
  resolution = 0.5, verbose = FALSE,
  algorithm = 4, # leiden
  group.singletons = FALSE
)
seu_singlet <- RunUMAP(
  seu_singlet,
  reduction = "pca", dims = dims
)

5.

dims = 1:10

seu_singlet <- FindNeighbors(
  seu_singlet,
  reduction = "pca", dims = dims,
  k.param = 30,
  annoy.metric = "manhattan"
)
seu_singlet <- FindClusters(
  seu_singlet,
  resolution = 0.5, verbose = FALSE,
  algorithm = 4, # leiden
  group.singletons = FALSE
)
seu_singlet <- RunUMAP(
  seu_singlet,
  reduction = "pca", dims = dims
)

[florekem]

6.

dims = 1:15
seu_singlet <- FindNeighbors(
  seu_singlet,
  reduction = "pca", dims = dims,
  k.param = 30,
  annoy.metric = "manhattan"
)
seu_singlet <- FindClusters(
  seu_singlet,
  resolution = 0.5, verbose = FALSE,
  algorithm = 4, # leiden
  group.singletons = FALSE
)
seu_singlet <- RunUMAP(
  seu_singlet,
  reduction = "pca", dims = dims
)

[florekem]

1.1.

RDS/seu_singlet-clustered-27nov2024.rds

An object of class Seurat 
64850 features across 15250 samples within 4 assays 
Active assay: SCT (26200 features, 3000 variable features)
 3 layers present: counts, data, scale.data
 3 other assays present: RNA, HTO, ADT
 2 dimensional reductions calculated: pca, umap

dims <- 1:20
seu_singlet <- FindNeighbors(
  seu_singlet,
  reduction = "pca", dims = dims,
  k.param = 30
)
seu_singlet <- FindClusters(
  seu_singlet,
  resolution = 0.8, verbose = FALSE,
  algorithm = 4, # leiden
  group.singletons = FALSE
)
seu_singlet <- RunUMAP(
  seu_singlet,
  reduction = "pca", dims = dims
)

(it may be worth to remove them, but not sure, not too many of them)