add cdr metadata

download_oas_data.py

@@ -2,22 +2,18 @@ import pandas as pd
 
 # Download data from OAS
 list_of_df = []
-with open("scripts/oas_uri_list.txt", "r") as f:
+with open("oas_uri_list.txt", "r") as f:
     for url in f:
         print(url)
         df = pd.read_csv(url, skiprows=1, compression="gzip")
         df = df[df["complete_vdj_heavy"] == "T"]
         df = df[df["complete_vdj_light"] == "T"]
-        df = df[["sequence_alignment_aa_heavy", "sequence_alignment_aa_light"]]
-        df["sequence_alignment_aa_joint"] = (
-            df["sequence_alignment_aa_heavy"]
-            + "-" * 25
-            + df["sequence_alignment_aa_light"]
-        )
+        df = df[["sequence_alignment_aa_heavy", "cdr1_start_heavy", "cdr1_end_heavy", "cdr2_start_heavy", "cdr2_end_heavy", "cdr3_start_heavy", "cdr3_end_heavy", "sequence_alignment_aa_light", "cdr1_start_light", "cdr1_end_light", "cdr2_start_light", "cdr2_end_light", "cdr3_start_light", "cdr3_end_light"]]
         list_of_df.append(df)
 df = pd.concat(list_of_df, ignore_index=True)
 train=df.sample(frac=0.8, ignore_index=True)
 test = df.drop(train.index)
 
-train.to_parquet("data/train.parquet", index=False)
-test.to_parquet("data/test.parquet", index=False)
+train.to_parquet("train.parquet", index=False)
+test.to_parquet("test.parquet", index=False)
+

test.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:bf99e45052397991e961d39868008b0e5c1b44a55979aacb5c95158481cf26eb
-size 21958383
+oid sha256:35058619ee7d0dc9b57966825ea075056b3e7fde75d9013dda0a8dbee6cfe076
+size 12091279

train.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:512307b54801f79503a4e4a642ea15351e93b081a5895148f01ceb71bedaab29
-size 77878172
+oid sha256:eeb084b4aaaa46a627e98b730eb70f86bcae1185f65b610a5f982ad3bb1a50e3
+size 43728897