ascend .5

.gitignore

@@ -336,3 +336,4 @@ tags
 .renku.lock
 .renku/tmp
 .renku/cache
+.Rproj.user

README.md

@@ -7,6 +7,10 @@
 
  Ascend clustering method from https://academic.oup.com/gigascience/article/8/8/giz087/5554286  
 
+Problem with newest version fom https://github.com/powellgenomicslab/ascend
+
+runPCA doesn't work, even on their own example dataset:  `unable to find an inherited method for function ‘int_colData’ for signature ‘"NULL"’`
+
 ## Method template 
 
 This template will help you to add a method to an omnibenchmark project. For each method that you have, **one dedicated repository** has to be created from this template to upload the method on the renku system. 

install.R

@@ -18,6 +18,6 @@ bioconductor_packages <- c("BiocParallel", "BiocGenerics",
                            "GenomeInfoDbData")
 BiocManager::install(bioconductor_packages)
 
-devtools::install_github("powellgenomicslab/ascend")
-
+# devtools::install_github("powellgenomicslab/ascend") #final version
+devtools::install_github("IMB-Computational-Genomics-Lab/ascend", ref = "b415c0918f10306829b64bfad37f5d08da049ef7") #v0.5 as in Duo et al.
 

src/ascend-clustering.R

@@ -2,7 +2,16 @@
 ##  ------------------------------ Ascend clustering ------------------------------------------ ##
 ### ------------------------------------------------------------------------------------------ ###
 
+
+# filtered_rawcounts = "data/filter_expression/filter_expression_koh_filtered_counts.mtx.gz"
+# 
+# meta_file = "data/filter_expression/koh_ttc_meta.json"
+# 
+# cluster_res = "data/ascend-clustering/ascend-clustering_filter_expression_1d45b_koh__k_2__seed_12876_cluster_res.json"
 # 
+# seed = 12876
+# 
+# k = 2
 
 ### ------------------------------------------------------------------------------------------ ###
 ##  -------------------------------------- Libraries ------------------------------------------ ##
@@ -18,6 +27,7 @@ suppressPackageStartupMessages({
   library(ascend)
   library(BiocParallel)
   library(class)
+  library(dplyr)
 })
 
 ### ------------------------------------------------------------------------------------------ ###
@@ -73,18 +83,33 @@ colnames(dat) <- meta$cell_id
 ##  -------------------------------------- Method --------------------------------------------- ##
 ### ------------------------------------------------------------------------------------------ ###
 
+rowdat <- data.frame("gene_names" = rownames(dat))
+a <- SingleCellExperiment(list(counts = dat))
+em_set <- EMSet(a, 
+                colInfo = meta,
+                rowInfo = rowdat)
+normset <- normaliseByRLE(em_set)
+pcaset <- runPCA(normset) ##
+pcaset <- ReduceDimensions(pcaset, n = params$nPC)
+
+library(ascend)
+data(raw_set)
+EMSet <- raw_set
 
 set.seed(seed)
-sce <- SingleCellExperiment(assays = list(counts = as.matrix(dat)))
-dat <- counts(sce)
-emset <- NewEMSet(ExpressionMatrix = dat)
-normset <- NormaliseByRLE(emset)
-pcaset <- RunPCA(normset)
+colnames(dat) <- meta$cell_id
+rownames(dat) <- paste0("gene", 1:nrow(dat))
+meta$cell_barcode <- paste0("cell:", 1:ncol(dat))
+meta <- meta %>%
+  select(cell_barcode, everything())
+emset <- EMSet(x = dat)
+normset <- normaliseByRLE(emset)
+pcaset <- runPCA(normset) ##
 pcaset <- ReduceDimensions(pcaset, n = params$nPC)
 ## Try to cluster all cells. If it fails (if there are outliers), cluster
 ## the ones that can be clustered and assign the remaining ones to a
 ## cluster with kNN on the extracted PCs.
-resset <- RunCORE(pcaset, conservative = TRUE, nres = 40, 
+resset <- runCORE(pcaset, conservative = TRUE, nres = 40, 
                 remove_outlier = TRUE)
 ## Select the height to use
 ks <- resset@Clusters$KeyStats