Use seed for nondeterministic processes vib-singlecell-nf/vsn-pipelin…

…es#43
vib-singlecell-nf · Feb 14, 2020 · 9379696 · 9379696
1 parent 1abc146
commit 9379696
Show file tree

Hide file tree

Showing 5 changed files with 31 additions and 7 deletions.
diff --git a/bin/dim_reduction/sc_dim_reduction.py b/bin/dim_reduction/sc_dim_reduction.py
@@ -38,7 +38,7 @@
 )
 
 parser.add_argument(
-    "-s", "--svd-solver",
+    "-v", "--svd-solver",
     type=str,
     action="store",
     dest="svd_solver",
@@ -72,6 +72,15 @@
     help="Use the MulticoreTSNE package by D. Ulyanov if it is installed."
 )
 
+parser.add_argument(
+    "-s", "--seed",
+    type=int,
+    action="store",
+    dest="seed",
+    default=0,
+    help="Use this integer seed for reproducibility."
+)
+
 args = parser.parse_args()
 
 # Define the arguments properly

diff --git a/bin/nn/sc_neighborhood_graph.py b/bin/nn/sc_neighborhood_graph.py
@@ -37,6 +37,16 @@
     help="Use this many PCs. If n_pcs==0 use .X if use_rep is None."
 )
 
+parser.add_argument(
+    "-s", "--seed",
+    type=int,
+    action="store",
+    dest="seed",
+    default=0,
+    help="Use this integer seed for reproducibility."
+)
+
+
 args = parser.parse_args()
 
 # Define the arguments properly
@@ -57,7 +67,8 @@
 sc.pp.neighbors(
     adata=adata,
     n_neighbors=args.n_neighbors,
-    n_pcs=args.n_pcs
+    n_pcs=args.n_pcs,
+    random_state=args.seed
 )
 
 # I/O

diff --git a/processes/dim_reduction.nf b/processes/dim_reduction.nf
@@ -94,7 +94,8 @@ process SC__SCANPY__DIM_REDUCTION {
 		tuple \
 			val(sampleId), \
 			path("${sampleId}.SC__SCANPY__DIM_REDUCTION_${method}.${!isParamNull(stashedParams) ? uuid + '.' : ''}${processParams.off}"), \
-			val(stashedParams)
+			val(stashedParams), \
+			val(nComps)
 
 	script:
 		def sampleParams = params.parseConfig(sampleId, params.global, params.sc.scanpy.dim_reduction.get(params.method))
@@ -110,6 +111,7 @@ process SC__SCANPY__DIM_REDUCTION {
 		_processParams.setConfigParams(processParams)
 		"""
 		${binDir}dim_reduction/sc_dim_reduction.py \
+			${'--seed ' + (params.global.containsKey('seed') ? params.global.seed: params.seed)} \
 			--method ${processParams.dimReductionMethod} \
 			${(processParams.containsKey('svdSolver')) ? '--svd-solver ' + processParams.svdSolver : ''} \
 			${(processParams.containsKey('nNeighbors')) ? '--n-neighbors ' + processParams.nNeighbors : ''} \

diff --git a/processes/neighborhood_graph.nf b/processes/neighborhood_graph.nf
@@ -88,8 +88,9 @@ process SC__SCANPY__NEIGHBORHOOD_GRAPH {
 	output:
         tuple \
             val(sampleId), \
-            path("${sampleId}.SC__SCANPY__NEIGHBORHOOD_GRAPH.${processParams.off}") \
-            val(stashedParams)
+            path("${sampleId}.SC__SCANPY__NEIGHBORHOOD_GRAPH.${processParams.off}"), \
+            val(stashedParams),
+			val(nComps)
 
 	script:
         def sampleParams = params.parseConfig(sampleId, params.global, params.sc.scanpy.filter)
@@ -103,9 +104,10 @@ process SC__SCANPY__NEIGHBORHOOD_GRAPH {
 		_processParams.setEnv(this)
 		_processParams.setConfigParams(processParams)
         """
-        ${binDir}cluster/sc_neighborhood_graph.py \
+        ${binDir}nn/sc_neighborhood_graph.py \
             $f \
             ${sampleId}.SC__SCANPY__NEIGHBORHOOD_GRAPH.${processParams.off} \
+			${'--seed ' + (params.global.containsKey('seed') ? params.global.seed: params.seed)} \
             ${(processParams.containsKey('nNeighbors')) ? '--n-neighbors ' + processParams.nNeighbors : ''} \
 			${_processParams.getNCompsAsArgument(nComps)}
         """

diff --git a/workflows/cluster_identification.nf b/workflows/cluster_identification.nf
@@ -3,7 +3,7 @@ nextflow.preview.dsl=2
 //////////////////////////////////////////////////////
 //  process imports:
 
-include '../../utils/processes/utils.nf'
+include '../../utils/processes/utils.nf' params(params)
 
 // scanpy:
 include '../processes/cluster' params(params)