Merge pull request #42 from sansomlab/yz_singleR

YZ_general

R/scripts/singleR_run.R

@@ -44,7 +44,7 @@ ref_path = file.path(opt$refstashdir, paste(opt$reference,"rds",sep="."))
 print(ref_path)
 ref.se <- readRDS(ref_path)
 
-if (grepl("ENSMUSG", rownames(ref.se)[1]))
+if (grepl("ENSMUSG", rownames(ref.se)[1]) & grepl("ENSG",rownames(sce)[1]))
 {
   message("cleaning up mouse ensembl indentifiers")
   rownames(ref.se) <- gsub("ENSMUSG", "ENSG", rownames(ref.se))

cellhub/pipeline_cluster.py

@@ -154,7 +154,7 @@ def taskSummary(infile, outfile):
     tab = pd.DataFrame(list(zip(tasks,run)),columns=["task","run"])
     tab.to_latex(buf=outfile, index=False, escape=True)
 
-
+    
 @files(PARAMS["source_anndata"],
        "preflight.sentinel")
 def preflight(infile, outfile):
@@ -428,20 +428,6 @@ def compareClusters(infile, outfile):
 
     reductiontype = PARAMS["source_rdim_name"]
 
-    if t.resolution == "predefined":
-
-        # TODO: Fix.
-        cluster_file = sample + ".cluster_ids.rds"
-
-        if os.path.exists(cluster_file):
-            predefined = "--predefined=%(cluster_file)s" % locals()
-
-        else:
-            raise ValueError("Predefined cluster assignment file (%(cluster_file)s) not found" % locals())
-
-    else:
-        predefined = ""
-
     statement = '''python %(cellhub_code_dir)s/python/cluster_compare.py
                    --source_anndata=%(source_anndata)s
                    --clusterids=%(cluster_ids)s
@@ -1911,15 +1897,13 @@ def cellxgene(infile, outfile):
 # ##################### full target: to run all tasks ####################### #
 # ########################################################################### #
 
-
 @follows(report, cellxgene)
 def full():
     pass
 
 
 # ------------------- < ***** end of pipeline **** > ------------------------ #
 
-
 def main(argv=None):
     if argv is None:
         argv = sys.argv
@@ -1929,4 +1913,4 @@ def main(argv=None):
 if __name__ == "__main__":
     sys.exit(P.main(sys.argv))
 
-#endregion
+#endregion

cellhub/pipeline_fetch_cells.py

@@ -144,7 +144,7 @@ def GEX(infile, outfile):
     TODO: support down-sampling
     '''
 
-    t = T.setup(infile, outfile, PARAMS, memory=PARAMS["resource_memory"])
+    t = T.setup(infile, outfile, PARAMS, memory=PARAMS["resources_memory"])
 
     cell_table = infile.replace(".sentinel", ".tsv.gz")
     api_path = os.path.join(PARAMS["cellhub_location"],"api")
@@ -177,7 +177,7 @@ def ADT(infile, outfile):
     TODO: support down-sampling
     '''
 
-    t = T.setup(infile, outfile, PARAMS, memory=PARAMS["resource_memory"])
+    t = T.setup(infile, outfile, PARAMS, memory=PARAMS["resources_memory"])
 
     cell_table = infile.replace(".sentinel", ".tsv.gz")
     api_path = os.path.join(PARAMS["cellhub_location"],"api")

cellhub/tasks/celldb.py

@@ -49,6 +49,7 @@ def load(table_name,
         statement = '''python -m cgatcore.tables
                                --regex-filename "%(regex_filename)s"
                                --cat "filename"
+                               --missing-value NULL
                                %(table_path)s/%(glob)s
                        | gzip -c
                        > %(table_file)s

cellhub/yaml/pipeline_cluster.yml

@@ -114,7 +114,8 @@ runspecs:
 # --------------------------
 neighbors:
 
-  # scanpy | hnsw
+  # scanpy | hnsw | precomputed
+  # if precomputed -> adata file needs to have neighbors to read 
   method: hnsw
 
   # for hnsw
@@ -502,4 +503,4 @@ cellxgene:
   # only one factor may be specified
   # default: None
   cluster_split: None
-
+

python/cluster_neighbor_graph.py

@@ -36,7 +36,7 @@
 parser.add_argument("--ncomps", default=20, type=int,
                     help="Number of reduced components to include in the knn computation")
 parser.add_argument("--method", default="scanpy", type=str,
-                    help="scanpy|hnsw|sklearn (scanpy uses pynndescent)")
+                    help="scanpy|hnsw|sklearn|precomputed (scanpy uses pynndescent)")
 parser.add_argument("--k", default=20, type=int,
                     help="number of neighbors")
 parser.add_argument("--metric", default="euclidean", type=str,
@@ -87,7 +87,6 @@
               use_rep = nn_rdims)
 
 
-
 elif args.method == "hnsw":
 
     L.info("Computing neighbors using hnswlib (with scvelo a la pegasus!)")
@@ -129,6 +128,22 @@
               # metric_kwds = {}
               num_threads=args.threads)
 
+
+elif args.method == "precomputed":
+
+   L.info("Using precomputed neighbors")
+
+   from scanpy._utils import NeighborsView
+
+   #add knn to adata
+   adata.uns["neighbors"] = sourceAdata.uns["neighbors"]
+   adata.obsp = sourceAdata.obsp
+
+   #rename rep_use to 'X___'
+   neighbors = NeighborsView(adata, "neighbors")
+   neighbors["params"]["use_rep"]='X_'+args.reduced_dims_name+'_copy_for_nn' 
+   adata.uns["neighbors"]["params"]=neighbors["params"] 
+
 else:
     raise ValueError("nn method not recognised")
 

python/fetch_cells_from_h5.py

@@ -42,6 +42,7 @@
 # ########################################################################### #
 
 L.info("Reading cell table")
+
 cell_table = pd.read_csv(args.cells, sep="\t")
 
 good_cols = [x for x in cell_table.columns if
@@ -55,9 +56,6 @@
                     zip([x.split("-")[0] for x in cell_table["barcode"].values],
                         cell_table["library_id"].values)]  
 
-print(cell_table)
-print(cell_table.index)
-
 libraries = set(cell_table["library_id"].values)
 
 anndata_slices = {}
@@ -158,7 +156,7 @@
 
 # drop unnecessary .obs columns
 anndata.obs.drop("barcode", axis=1, inplace=True)
-
+    
 # save the anndata
 anndata.write_h5ad(os.path.join(args.outdir, args.outname))