fix

downloads-generation/data_curated/GENERATE.sh

@@ -40,7 +40,10 @@ do
     CURATE_BY_PMID_ARGS+=$(echo --item $pmid raw/$pmid/* ' ')
 done
 
-time python curate_by_pmid.py $CURATE_BY_PMID_ARGS --out curated.by_pmid.csv --debug
+time python curate_by_pmid.py $CURATE_BY_PMID_ARGS \
+    --out nontraining_curated.by_pmid.csv
+
+bzip2 nontraining_curated.by_pmid.csv
 
 rm -rf raw
 

downloads-generation/data_curated/curate_by_pmid.py

@@ -126,6 +126,12 @@ def handle_pmid_23481700(filename):
         results.append(result_df)
 
     result_df = pandas.concat(results, ignore_index=True)
+
+    # Rename samples to avoid a collision with the JY sample in PMID 25576301.
+    result_df.sample_id = result_df.sample_id.map({
+        "JY": "JY.2015",
+        "HHC": "HHC.2015",
+    })
     return result_df
 
 
@@ -666,6 +672,10 @@ def run():
     else:
         print("No nulls.")
 
+    # Each sample should be coming from only one experiment.
+    assert df.groupby("sample_id").pmid.nunique().max() == 1, (
+        df.groupby("sample_id").pmid.nunique().sort_values())
+
     df.to_csv(args.out, index=False)
     print("Wrote: %s" % os.path.abspath(args.out))
 

mhcflurry/downloads.yml

@@ -50,7 +50,7 @@ releases:
               default: false
 
             - name: data_curated
-              url: https://github.com/openvax/mhcflurry/releases/download/pre-1.4.0/data_curated.20190925.tar.bz2
+              url: https://github.com/openvax/mhcflurry/releases/download/pre-1.4.0/data_curated.20190927.tar.bz2
               default: true
 
             # Older downloads