Questions about Cluster identification

[FionaMoon]

Hi , I have some questions about identifyClusters function.

1. I can get the clustering result by using get_clusters , but I can't find which slot in SingCellaR object contains the infomation about clusters by using str(SingCellaR object).
2. People always try different resolution to find the proper clustering result when using Seurat, just like this

   data <- data %>% 
   FindNeighbors(dims = 1:15) %>% 
   FindClusters(resolution = c(seq(0,2,0.2)))
   # Then using clustree to see changes of clusters
   pdf(file = "./Seurat/clustree.pdf",width = 12, height = 12)
   clustree(data, prefix = "SCT_snn_res.", node_label= "percent.mt",
        node_label_aggr = "min",  node_label_size = 2)
   dev.off()

   So should I change parameters in identifyClusters function to find the proper clustering result ? Thank you in advance. Fiona

[supatt-lab]

Hi,

1. clustering information is stored in the slot @sc.clusters
2. SingCellaR uses the louvain clustering function from 'igraph::cluster_louvain'. So, it doesn’t have a ‘resolution’ parameter. You can play around with ’n.dims.use,’ n.neighbors’, and ‘knn.metric’ parameters to see the differences (see below). SingCellaR also provides K-means clustering function 'clustering_KMeansOnKNN_Graph'. You can try to see if this helps.

identifyClusters( object, dim_reduction_method = c("pca", "nnmf", "lsi"), useIntegrativeEmbeddings = FALSE, integrative_method = c("combat", "seurat", "harmony", "supervised_harmony"), n.dims.use = 30, n.neighbors = 30, knn.n_trees = 50, knn.metric = c("euclidean", "cosine"), knn.n_threads = 1 )

[FionaMoon]

Thank you, you answer really helps !