Merge pull request #7 from til-unc/add-human-allele-aliases

Copied human allele aliases from IMGT
pirl-unc · Nov 23, 2020 · b5d8581 · b5d8581
2 parents 9988a9a + 35afb7e
commit b5d8581
Show file tree

Hide file tree

Showing 19 changed files with 661 additions and 198 deletions.
diff --git a/mhcgnomes/allele.py b/mhcgnomes/allele.py
@@ -73,7 +73,7 @@ def num_allele_fields(self):
 
     @property
     def name(self):
-        return ":".join(self.allele_fields)
+        return ":".join(self.allele_fields) + "".join(self.annotations)
 
     def restrict_num_allele_fields(
             self,

diff --git a/mhcgnomes/allele_without_gene.py b/mhcgnomes/allele_without_gene.py
@@ -40,7 +40,7 @@ def to_string(
         """
         Return allele strings like "BoLA-T2C"
         """
-        species_str = self.species_str.to_string(
+        species_str = self.species.to_string(
             include_species=include_species,
             use_old_species_prefix=use_old_species_prefix)
         return "%s-%s" % (species_str, self.name)

diff --git a/mhcgnomes/common.py b/mhcgnomes/common.py
@@ -19,7 +19,14 @@ def unique(xs : Iterable):
     """
     if type(xs) is list and len(xs) == 0 or len(xs) == 1:
         return xs
-    return list(set(xs))
+    result = []
+    unique_set = set()
+    for xi in xs:
+        if xi in unique_set:
+            continue
+        result.append(xi)
+        unique_set.add(xi)
+    return result
 
 def arg_to_cache_key(x):
     if type(x) in {list, tuple}:
@@ -49,14 +56,17 @@ def cached_fn(*args, **kwargs):
         return cache[key]
     return cached_fn
 
-def normalize_string(name, chars_to_remove="-_'"):
+def normalize_string(name, chars_to_remove="-_':"):
     """
     Return uppercase string without any surrounding whitespace and
-    without any characters such as '-', '_' or "'"
+    without any characters such as '-', '_' ':' or "'"
     """
+    if name is None:
+        return None
     if type(name) in (float, int):
         name = str(name)
 
+
     if not isinstance(name, str):
         return name
 

diff --git a/mhcgnomes/data.py b/mhcgnomes/data.py
@@ -60,6 +60,12 @@ def load(
     normalize_first_level_keys=True,
     normalize_second_level_keys=True)
 
+# Dictionary mapping species -> gene -> list of known allele names
+known_alleles = load(
+    "known_alleles.yaml",
+    normalize_first_level_keys=True,
+    normalize_second_level_keys=True)
+
 # Dictionary mapping species to haplotype name to list of alleles
 haplotypes = load(
     "haplotypes.yaml",

diff --git a/mhcgnomes/data/allele_aliases.yaml b/mhcgnomes/data/allele_aliases.yaml
@@ -8,66 +8,34 @@
 ####################################################
 #
 # Swine allele aliases manually scraped from:
-# https://www.ebi.ac.uk/ipd/mhc/
+# - https://www.ebi.ac.uk/ipd/mhc/
 #
 ####################################################
 SLA:
   # Allele names from NetMHCpan
   # many seem to to be derived from Chinese publications that I can't read,
-  #
-  1*CHANGDA: ~
   1*CDY.AA: 1*CDY
   2*CDY.AA: 2*CDY
   3*CDY.AA: 3*CDY
-  1*CDY: ~
-  2*CDY: ~
-  3*CDY: ~
-  1*HB01: ~
-  1*HB02: ~
-  1*HB03: ~
-  1*HB04: ~
-  2*HB01: ~
-  3*HB01: ~
   1*LWH.AA: 1*LWH
   2*LWH.AA: 2*LWH
   3*LWH.AA: 3*LWH
-  1*LWH: ~
-  2*LWH: ~
-  3*LWH: ~
   1*TPK.AA: 1*TPK
   2*TPK.AA: 2*TPK
   3*TPK.AA: 3*TPK
-  1*TPK: ~
-  2*TPK: ~
-  3*TPK: ~
   1*YC.AA: 1*YC
   2*YC.AA: 2*YC
   3*YC.AA: 3*YC
-  1*YC: ~
-  2*YC: ~
-  3*YC: ~
   1*YDL.AA: 1*YDL
   2*YDL.AA: 2*YDL
   3*YDL.AA: 3*YDL
-  1*YDL: ~
-  2*YDL: ~
-  3*YDL: ~
-  1*YDL01: ~
-  2*YDL02: ~
   1*YDY.AA: 1*YDY
   2*YDY.AA: 2*YDY
   3*YDY01.AA: 3*YDY01
   3*YDY02.AA: 3*YDY01
-  1*YDY: ~
-  2*YDY: ~
-  3*YDY01: ~
-  3*YDY02: ~
   1*YTH.AA: 1*YTH
   2*YTH.AA: 2*YTH
   3*YTH.AA: 3*YTH
-  1*YTH: ~
-  2*YTH: ~
-  3*YTH: ~
 
   # provisional allele names mapped to their permanent names
   1*01rh28: 1*01:03
@@ -289,9 +257,9 @@ SLA:
   DRB3*01bL85: DRB3*05:01
 Patr:
   # footnote in Table 1 of 10.1016/j.molimm.2009.09.003
-  DRB*W901: DRB3*0701
-  DRB*W902: DRB3*0702
-  DRB*W903: DRB3*0703
+  DRB*W901: DRB3*07:01
+  DRB*W902: DRB3*07:02
+  DRB*W903: DRB3*07:03
 BoLA:
   #  Source: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4225172/
   HD6: 6*013:01
@@ -300,8 +268,196 @@ BoLA:
   AW10: 3*001:01
   T2a: 2*012:01
   T2b: 6*041:01
-  T2c: ~
-  T7: ~
-  T5: ~
   # TODO: What to do with the invariant DR alpha genes of many species?
   # DRA: DRA*01:01
+####################################################
+#
+# Human deleted and renamed alleles copied from:
+# - https://www.ebi.ac.uk/cgi-bin/ipd/imgt/hla/deleted.cgi
+#
+####################################################
+HLA:
+  # renamed to entirely different alleles
+  A*01:05N: A*01:04:01:01N
+  A*01:34N: A*01:01:38L
+  A*02:01:16: A*02:134
+  A*02:01:20: A*02:01:18
+  A*02:23: A*02:22:01
+  A*02:98: A*02:96
+  A*02:01:82: A*02:01:84
+  A*02:17:01: A*02:17:02:01
+  A*03:194: A*03:213
+  A*03:200Q: A*03:266N
+  A*03:260: A*03:284N
+  A*11:28: A*11:15:02
+  A*11:53: A*11:02:01:01
+  A*23:69: A*23:01:01
+  A*24:12: A*24:08
+  A*24:16: A*31:08
+  A*24:65: A*24:13:02
+  A*24:211: A*24:135:02
+  A*26:03:02: A*26:111
+  A*26:44: A*26:43:02
+  A*30:05: A*30:04:01:01
+  A*30:21: A*30:11:02
+  A*31:011: A*31:01:02
+  A*33:02: A*33:03:01
+  A*33:38: A*33:44
+  B*08:06: B*08:20:01
+  B*13:05: B*13:04
+  B*13:24: B*13:22:02
+  B*15:01:05: B*15:120
+  B*15:22: B*35:43
+  B*15:41: B*15:39:01:01
+  B*15:59: B*35:44
+  B*18:16: B*18:14
+  B*27:051: B*27:05:02
+  B*27:22: B*27:06:01:01
+  B*27:215N: B*27:05:02:01
+  B*35:73: B*35:08:03
+  B*35:43:02: B*35:185
+  B*39:012: B*39:01:01:01
+  B*39:21: B*39:24
+  B*40:17: B*40:16:01:01
+  B*40:41: B*40:40
+  B*44:01: B*44:02:01:01
+  B*45:26: B*45:01:01:01
+  B*47:01:01:01: B*47:01:01:03
+  B*49:15: B*49:01:01
+  B*50:03: B*50:02:01:01
+  B*51:25: B*51:22
+  B*51:47: B*51:09:02
+  B*55:06: B*55:04
+  B*56:55:01:01: B*56:55:01:02
+  B*58:30: B*58:01:01
+  B*78:01:01:01: B*78:01:01:02
+  B*79:01: B*15:18:01
+  B*95:30: B*15:27:02
+  C*01:63:01: C*01:141
+  C*02:02:05: C*02:10:06
+  C*02:16:01: C*02:137
+  C*03:12: C*03:19
+  C*03:99:01: C*01:169:01
+  C*03:99:02: C*01:169:02
+  C*04:01:01:30: C*04:01:01:27
+  C*08:64: C*08:33:03
+  C*12:274:02N: C*12:329N
+  C*12:299N: C*12:03:01:01
+  C*15:20: C*15:27
+  C*17:01:01:01: C*17:01:01:02
+  C*01:01: C*01:02:01
+  C*02:01: C*02:02:02
+  C*02:02:04: C*02:10
+  C*02:16:03: C*02:16:02
+  C*03:01: C*03:04:01:01
+  C*04:02: C*04:01:01:01
+  C*04:21: C*04:15:02
+  C*04:22: Cw*04:21
+  C*05:02: C*05:09
+  C*06:01: C*06:02:01:01
+  C*06:02:02: C*06:17
+  C*07:34: C*07:27:02
+  C*12:01: C*12:02:02
+  C*14:01: C*14:02:01
+  C*15:01: C*15:02:01
+  C*15:14: C*15:10:02
+  C*16:03: C*14:03:01:01
+  C*16:042: C*16:04:01
+  C*16:05: C*16:04:01
+  DPA1*01:01: DPA1*01:03:01
+  DPA1*01:02: DPA1*01:03:01
+  DPA1*02:02:01: DPA1*02:07:01
+  DPA1*02:02:02:06Q: DPA1*02:02:02:01
+  DPB1*02:011: DPB1*02:01:02
+  DPB1*35:01:02: DPB1*621:01
+  DPB1*42:01: DPB1*31:01:01:01
+  DPB1*43:01: DPB1*28:01
+  DPB1*793:01N: DPB1*786:01:02N
+  DQA1*01:01:01:04: DQA1*01:01:01:01
+  DQA1*02:06: DQA1*02:01:01:01
+  DQA1*03:012: DQA1*03:02
+  DQA1*03:01:01:02: DQA1*03:01:01:01
+  DQA1*03:02:01:03: DQA1*03:02:01:02
+  DQA1*05:013: DQA1*05:05
+  DQB1*02:02:06:01: DQB1*02:02:06:02
+  DQB1*03:031: DQB1*03:03:02
+  DQB1*03:01:01:13: DQB1*03:01:01:07
+  DQB1*03:325: DQB1*03:317:02
+  DQB1*04:02:01:02: DQB1*04:02:01:01
+  DQB1*04:02:01:03: DQB1*04:02:01:01
+  DQB1*06:220: DQB1*06:217
+  DRB1*03:11:02: DRB1*03:81
+  DRB1*04:05:12: DRB1*04:05:14
+  DRB1*04:94:02N: DRB1*04:212N
+  DRB1*07:02: DRB1*07:01:01:01
+  DRB1*08:031: DRB1*08:03:02
+  DRB1*08:01:03: DRB1*08:01:01
+  DRB1*09:011: DRB1*09:01:02
+  DRB1*11:71: DRB1*11:02:01:01
+  DRB1*11:11:02: DRB1*11:11:01
+  DRB1*12:031: DRB1*12:01:01
+  DRB1*14:66: DRB1*14:32:02
+  DRB1*16:06: DRB1*16:05:01
+  DRB3*01:01:01: DRB3*01:01:02:01
+  DRB4*01:011:02N: DRB4*01:03:01:02N
+  DRB5*02:01: DRB5*02:02
+  E*01:02: E*01:01:01:01
+  E*01:21:02N: E*01:117N
+  F*01:01:03:01: F*01:01:01:09
+  F*01:01:03:02: F*01:01:01:10
+  F*01:01:03:03: F*01:01:01:11
+  F*01:01:03:04: F*01:01:01:12
+  G*01:01:10: G*01:04:04
+  MICA*007:02: MICA*111:01:01
+  MICA*007:04: MICA*112:01:01
+  MICA*008:07: MICA*008:01:10
+  MICA*021: MICA*012:03
+  MICA*071: MICA*017
+  MICB*017: MICB*005:05
+
+  # alleles with suffix changes
+  A*01:159: A*01:159Q
+  A*01:281: A*01:281Q
+  A*01:301: A*01:301Q
+  A*02:01:14: A*02:01:14Q
+  A*02:437: A*02:437Q
+  A*02:581: A*02:581Q
+  A*02:728: A*02:728Q
+  A*02:795: A*02:795Q
+  A*11:50: A*11:50Q
+  A*11:52: A*11:52Q
+  A*23:19Q: A*23:19N
+  A*24:02:01:17: A*24:02:01:17Q
+  A*24:329: A*24:329Q
+  A*24:378: A*24:378Q
+  A*24:447: A*24:447Q
+  A*24:450: A*24:450Q
+  A*33:03:03: A*33:03:03Q
+  B*07:44: B*07:44N
+  B*13:08Q: B*13:08
+  B*18:01:01:12: B*18:01:01:12Q
+  B*18:106: B*18:106Q
+  B*27:185: B*27:185Q
+  B*3716: B*37:16Q
+  B*38:68Q: B*38:68L
+  B*41:56: B*41:56Q
+  B*44:02:01:13: B*44:02:01:13Q
+  B*44:142Q: B*44:142
+  B*44:160: B*44:160Q
+  C*02:67: C*02:67Q
+  C*03:23: C*03:23N
+  C*03:169: C*03:169Q
+  C*04:338: C*04:338Q
+  C*04:382: C*04:382Q
+  C*06:74: C*06:74Q
+  C*07:226: C*07:226Q
+  C*07:235: C*07:235Q
+  C*07:432: C*07:432Q
+  C*07:713: C*07:713Q
+  C*08:70: C*08:70Q
+  C*12:139: C*12:139Q
+  Cw*16:16: C*16:16Q
+  DRB4*01:14: DRB4*01:14N
+  DRB5*02:26: DRB5*02:26N
+  DQA1*01:07: DQA1*01:07Q
+  DPB1*939:01N: DPB1*939:01
diff --git a/mhcgnomes/data/known_alleles.yaml b/mhcgnomes/data/known_alleles.yaml
@@ -0,0 +1,45 @@
+SLA:
+  "1":
+    - CDY
+    - CHANGDA
+    - HB01
+    - HB02
+    - HB03
+    - HB04
+    - LWH
+    - TPK
+    - YC
+    - YTH
+    - YDY
+    - YDL
+    - YDL01
+  "2":
+    - CDY
+    - HB01
+    - LWH
+    - TPK
+    - YC
+    - YTH
+    - YDY
+    - YDL
+    - YDL02
+  "3":
+    - CDY
+    - HB01
+    - LWH
+    - TPK
+    - YC
+    - YTH
+    - YDY01
+    - YDY02
+    - YDL
+BoLA:
+  ~:
+    - HD6
+    - JSP.1
+    - D18.4
+    - AW10
+    - T2b
+    - T2c
+    - T5
+    - T7