adding new strains to include list and updating clades. fixed an erro…

…r in clade assignments to assign only new strains that are in the metadata
nextstrain · Feb 7, 2025 · 94d2a92 · 94d2a92
1 parent ab637d1
commit 94d2a92
Show file tree

Hide file tree

Showing 7 changed files with 2,131 additions and 21 deletions.
diff --git a/clade-labeling/h5n1-clades.tsv b/clade-labeling/h5n1-clades.tsv
diff --git a/clade-labeling/h5nx-clades.tsv b/clade-labeling/h5nx-clades.tsv
diff --git a/clade-labeling/update-clades.py b/clade-labeling/update-clades.py
@@ -27,34 +27,42 @@
 sequences = args.sequences
 subtype = args.subtype
 
-def find_new_strains(clades_file, new_strains):
-    old_strains = pd.read_csv(clades_file, sep="\t")['name']
-    new_strains = pd.read_csv(new_strains, sep="\t")['strain']
-
+def find_new_strains(clades_file, metadata):
+    old_strains = pd.read_csv(clades_file, sep="\t")['name'].tolist()
+    all_strains = pd.read_csv(metadata, sep="\t")['strain'].tolist()
+    new_strains = []
+
+    # if metadata strain not in old strains, return as new strain 
+    for a in all_strains: 
+        if a not in old_strains: 
+            new_strains.append(a)
+
+
     # union of the series -> combination of values present in either set 1 or set 2; so, sum of all elements
-    union = pd.Series(np.union1d(new_strains, old_strains))
-
-    # intersection of the series -> elements that are shared between the 2
-    intersect = pd.Series(np.intersect1d(new_strains, old_strains))
-
-    # uncommon elements in both the series; union - all elements of union that are in intersect
-    notcommonseries = union[~union.isin(intersect)]
-
-    new_strains = notcommonseries.tolist()
-
+#     union = pd.Series(np.union1d(all_strains, old_strains))
+#     
+#     # intersection of the series -> elements that are shared between the 2
+#     intersect = pd.Series(np.intersect1d(all_strains, old_strains))
+#   
+#     # uncommon elements in both the series; union - all elements of union that are in intersect
+#     notcommonseries = union[~union.isin(intersect)]
+#   
+    #new_strains = notcommonseries.tolist()
+    print(len(new_strains))
+    #print(new_strains)
     return(new_strains)
-
+    
 
 
 def separate_new_strains(new_strains_list, input_fasta, output_fasta):
     with open(output_fasta, "w") as outfile: 
         outfile.write("")
 
     for seq in SeqIO.parse(input_fasta, "fasta"):
-        strain = seq.description.split("|")[0]
+        strain = seq.description.split("|")[0].strip()
         full_sequence_header = seq.description
 
-        if strain in new_strains_list:             
+        if strain in new_strains_list: 
             with open(output_fasta, "a") as outfile:
                 outfile.write(">" + full_sequence_header + "\n" + str(seq.seq) + "\n")
 
@@ -82,8 +90,8 @@ def append_new_clades(new_clades, old_clades):
 separate_new_strains(new_strains, sequences, new_strains_fasta)
 
 
-"""run label and reformat the output"""
-print("\nrunning LABEL to assign clades to", len(new_strains), "new strains of", subtype)
+# """run label and reformat the output"""
+# print("\nrunning LABEL to assign clades to", len(new_strains), "new strains of", subtype)
 os.system('flu-amd-label-2023-05-05/./LABEL -D {new_strains_fasta} {label_output} H5v2023'.format(new_strains_fasta=new_strains_fasta, label_output=label_output))
 os.system('python clade-labeling/check-LABEL-annotations.py --label_output {label_output}_final.txt --output {new_clades_file}'.format(new_clades_file=new_clades_file, label_output=label_output))
 

diff --git a/config/h5n1/include_strains_h5n1_2y.txt b/config/h5n1/include_strains_h5n1_2y.txt
@@ -39,6 +39,7 @@ A/Chicken/BC/FAV-0266-1/2024
 A/Turkey/BC/FAV-0267-1/2024
 A/Wisconsin/179/2024
 A/Iowa/124/2024
+A/England/480160/2025
 
 # cat outbreak 
 A/tiger/USA/038158001/2024

diff --git a/config/h5n1/include_strains_h5n1_all-time.txt b/config/h5n1/include_strains_h5n1_all-time.txt
@@ -58,4 +58,6 @@ A/British_Columbia/PHL-2032/2024
 A/Chicken/BC/FAV-0285-1/2024
 A/Chicken/BC/FAV-0268-1/2024
 A/Chicken/BC/FAV-0266-1/2024
-A/Turkey/BC/FAV-0267-1/2024
+A/Turkey/BC/FAV-0267-1/2024
+A/England/480160/2025
+
diff --git a/config/h5nx/include_strains_h5nx_2y.txt b/config/h5nx/include_strains_h5nx_2y.txt
@@ -38,6 +38,8 @@ A/Chicken/BC/FAV-0268-1/2024
 A/Chicken/BC/FAV-0266-1/2024
 A/Louisiana/12/2024
 A/Turkey/BC/FAV-0267-1/2024
+A/England/480160/2025
+
 
 
 # US dairy cattle outbreak
@@ -383,4 +385,9 @@ A/cat/USA/037985004/2024
 A/cat/USA/037985008/2024
 A/cat/USA/037986004/2024
 A/cat/USA/037986006/2024
-A/cat/California/038279001/2024
+A/cat/California/038279001/2024
+
+# important low paths 
+A/Ruddy Turnstone/Delaware/549/2024
+A/Ruddy Turnstone/Delaware/550/2024
+A/Laughing Gull/Delaware/554/2024
diff --git a/config/h5nx/include_strains_h5nx_all-time.txt b/config/h5nx/include_strains_h5nx_all-time.txt
@@ -49,4 +49,6 @@ A/Chicken/BC/FAV-0285-1/2024
 A/Chicken/BC/FAV-0268-1/2024
 A/Chicken/BC/FAV-0266-1/2024
 A/Turkey/BC/FAV-0267-1/2024
+A/England/480160/2025
+