interpretation on counts for krist testing

sherlock/interpret_counts.sh

@@ -23,3 +23,4 @@ fi
 cd chrombpnet
 echo $( timestamp ): "python $PWD/src/evaluation/interpret/interpret.py --genome=$reference_fasta --regions=$regions --output_prefix=$output_prefix --model_h5=$model_h5 --profile_or_counts counts" | tee -a $logfile
 python $PWD/src/evaluation/interpret/interpret.py --genome=$reference_fasta --regions=$regions --output_prefix=$output_prefix --model_h5=$model_h5 --profile_or_counts counts
+

sherlock/run_atac.py

@@ -10,26 +10,26 @@
 bias_model="/scratch/groups/akundaje/anusri/chromatin_atlas/reference/bias.h5"
 
 # choose the cet of accessions to run
-#encode_ids = [line.strip() for line in open("atac_files/cell_type.txt").readlines()]
+encode_ids = [line.strip() for line in open("atac_files/cell_type.txt").readlines()]
 #encode_ids = [line.strip() for line in open("tissue_type.txt").readlines()]
 #encode_ids = [line.strip() for line in open("atac_files/primary_cell_accession.txt").readlines()]
-encode_ids = [line.strip() for line in open("atac_files/invitro_accession.tsv").readlines()]
+#encode_ids = [line.strip() for line in open("atac_files/invitro_accession.tsv").readlines()]
 
 done = []
 undone = []
 
 # only 50 GPU jobs allowed on sherlock so limit this array - an run in batches of 50
-for encode_id in encode_ids:
+for encode_id in encode_ids[0:1]:
 	if encode_id in done:
 		continue
 	if encode_id in undone:
 		continue
 	print(output_dir+"/"+encode_id)
 	#command = "bash start_preprocessing.sh " + encode_id + " " + output_dir + " " + metadata_tsv + " " + oak_dir
 	#command = "bash start_gc_matching.sh " + encode_id + " " + output_dir + " " + metadata_tsv + " " + oak_dir
-	command = "bash start_modeling.sh " + encode_id + " " + output_dir + " " + metadata_tsv + " " + oak_dir  + " " + bias_model
+	#command = "bash start_modeling.sh " + encode_id + " " + output_dir + " " + metadata_tsv + " " + oak_dir  + " " + bias_model
 	#command = "bash start_interpretation.sh " + encode_id + " " + output_dir + " " + metadata_tsv + " " + oak_dir
 	#command = "bash start_modisco.sh " + encode_id + " " + output_dir + " " + metadata_tsv + " " + oak_modisco_dir
-	#command = "bash start_interpretation_counts.sh " + encode_id + " " + output_dir + " " + metadata_tsv + " " + oak_dir
+	command = "bash start_interpretation_counts.sh " + encode_id + " " + output_dir + " " + metadata_tsv + " " + oak_dir
 	print(command)
 	os.system(command)

sherlock/run_interpretation_counts.sh

@@ -7,7 +7,8 @@ oak_dir=$3
 
 reference_fasta=/scratch/groups/akundaje/anusri/chromatin_atlas/reference/hg38.genome.fa
 singularity exec --nv /home/groups/akundaje/anusri/simg/tf-atlas_gcp-modeling.sif nvidia-smi
-singularity exec --nv /home/groups/akundaje/anusri/simg/tf-atlas_gcp-modeling.sif bash interpret_counts.sh $reference_fasta $dir/preprocessing/downloads/peaks_no_blacklist.bed.gz $dir/interpret_counts_full/$experiment $dir/chrombpnet_model/chrombpnet_wo_bias.h5
+#singularity exec --nv /home/groups/akundaje/anusri/simg/tf-atlas_gcp-modeling.sif bash interpret_counts.sh $reference_fasta $dir/preprocessing/downloads/peaks_no_blacklist.bed.gz $dir/interpret_counts_full/$experiment $dir/chrombpnet_model/chrombpnet_wo_bias.h5
+singularity exec --nv /home/groups/akundaje/anusri/simg/tf-atlas_gcp-modeling.sif bash interpret_counts.sh $reference_fasta $dir/preprocessing/downloads/testing.bed $dir/interpret_counts_full/$experiment $dir/chrombpnet_model/chrombpnet_wo_bias.h5
 wait
 cp -r $dir/interpret_counts_full $oak_dir
 

sherlock/start_interpretation_counts.sh

@@ -5,11 +5,17 @@ output_dir=$2
 metadata_tsv=$3
 oak_dir=$4
 
-if [[ -d $output_dir/$experiment ]] ; then
-    if [[ -f $output_dir/$experiment/chrombpnet_model/chrombpnet_wo_bias.h5 ]] ; then
-        if [[ -f $output_dir/$experiment/interpret_counts_full/$experiment.counts_scores.h5 ]] ; then
+if [[ -d $oak_dir/$experiment ]] ; then
+    if [[ -f $oak_dir/$experiment/chrombpnet_model/chrombpnet_wo_bias.h5 ]] ; then
+        if [[ -f $oak_dir/$experiment/interpret_counts_full/$experiment.counts_scores.h5 ]] ; then
             echo "count interpretations already exist - skipping"
         else
+            if [[ ! -f $output_dir/$experiment/chrombpnet_model/chrombpnet_wo_bias.h5 ]] ; then
+                echo "copying model form oak"
+                mkdir $output_dir/$experiment/chrombpnet_model/
+                cp $oak_dir/$experiment/chrombpnet_model/chrombpnet_wo_bias.h5 $output_dir/$experiment/chrombpnet_model/
+
+            fi
             mkdir $output_dir/$experiment/interpret_counts_full/
             cores=1
             sbatch --export=ALL --requeue \
@@ -19,7 +25,7 @@ if [[ -d $output_dir/$experiment ]] ; then
                 --mem=80G \
                 -o $output_dir/$experiment/interpret_counts_full/interpret.log.o \
                 -e $output_dir/$experiment/interpret_counts_full/interpret.log.e \
-                run_interpretation_counts.sh $experiment $output_dir/$experiment/ $oak_dir/$experiment/chrombpnet_model/
+                run_interpretation_counts.sh $experiment $output_dir/$experiment/ $oak_dir/$experiment/
         fi
     else
         echo "skipping interpretation - model not found"