Merge pull request #39 from ShawHahnLab/for-38-igblast-ref-from-species

Default to all available references for a given species for IgBLAST

Author: ressy

CHANGELOG.md

@@ -4,6 +4,8 @@
 
 ### Added
 
+ * Automatic usage of all available references for a given species in `igblast`
+   command ([#39])
  * `identity` command for calculating pairwise identity between arbitrary
    queries and references ([#31], [#37])
  * Support for showing basic tree topology for Newick-format files in `show`
@@ -15,6 +17,7 @@
  * broken pipes (such as from `igseq something | something else`) are now
    handled gracefully ([#30])
 
+[#39]: https://github.com/ShawHahnLab/igseq/pull/39
 [#37]: https://github.com/ShawHahnLab/igseq/pull/37
 [#35]: https://github.com/ShawHahnLab/igseq/pull/35
 [#33]: https://github.com/ShawHahnLab/igseq/pull/33

igseq/igblast.py

@@ -70,6 +70,11 @@ def igblast(
     LOGGER.info("given colmap: %s", colmap)
     LOGGER.info("given extra args: %s", extra_args)
     LOGGER.info("given threads: %s", threads)
+    if species and not ref_paths:
+        # If only species is given, default to using all available reference
+        # sets for that species
+        ref_paths = [_fuzzy_species_match(species)]
+        LOGGER.info("inferred ref path: %s", ref_paths[0])
     attrs_list = vdj.parse_vdj_paths(ref_paths)
     for attrs in attrs_list:
         LOGGER.info("detected ref path: %s", attrs["path"])
@@ -104,12 +109,11 @@ def detect_organism(species_det, species=None):
         species = species_det.pop()
         LOGGER.info("detected species: %s", species)
     # match species names if needed
-    species_key = re.sub("[^a-z]", "", species.lower())
-    if species not in SPECIESMAP and species_key in SPECIESOTHER:
-        species_new = SPECIESOTHER[species_key]
+    species_new = _fuzzy_species_match(species)
+    if species_new != species:
         LOGGER.info(
-            "detected species as synonym: %s -> %s -> %s", species, species_key, species_new)
-        species = species_new
+            "detected species as synonym: %s -> %s", species, species_new)
+    species = species_new
     try:
         organism = SPECIESMAP[species]
     except KeyError as err:
@@ -118,6 +122,15 @@ def detect_organism(species_det, species=None):
     LOGGER.info("detected IgBLAST organism: %s", organism)
     return organism
 
+def _fuzzy_species_match(species):
+    """Fuzzy-match one of our species names"""
+    species_key = re.sub("[^a-z]", "", species.lower())
+    try:
+        return SPECIESOTHER[species_key]
+    except KeyError as err:
+        keys = str(SPECIESMAP.keys())
+        raise util.IgSeqError(f"species not recognized.  should be one of: {keys}") from err
+
 @contextmanager
 def run_igblast(
     db_dir, organism, query_path, threads=1, fmt_in=None, colmap=None, extra_args=None):

test_igseq/test_igblast.py

@@ -34,6 +34,24 @@ def test_detect_organism(self):
             igblast.detect_organism(["species1", "species2"], "rhesus"),
             "rhesus_monkey")
 
+    def test_detect_ref(self):
+        """Test that appropriate reference is selected based on species when needed."""
+        # Should be able to infer ref from exact species name or synonym
+        with self.assertLogs(level="INFO") as log_cm:
+            igblast.igblast(ref_paths=None, query_path="-", species="rhesus", dry_run=True)
+            self.assertTrue(any("inferred ref path: rhesus" in msg for msg in log_cm.output))
+        with self.assertLogs(level="INFO") as log_cm:
+            igblast.igblast(ref_paths=None, query_path="-", species="rhesus_monkey", dry_run=True)
+            self.assertTrue(any("inferred ref path: rhesus" in msg for msg in log_cm.output))
+        # Shouldn't infer ref when explicitly given
+        with self.assertLogs(level="INFO") as log_cm:
+            igblast.igblast(ref_paths=["rhesus"], query_path="-", species="rhesus", dry_run=True)
+            self.assertFalse(any("inferred ref path: rhesus" in msg for msg in log_cm.output))
+        # Should still catch unknown names
+        with self.assertRaises(IgSeqError) as err_cm:
+            igblast.igblast(ref_paths=None, query_path="-", species="unknown", dry_run=True)
+            self.assertIn("species not recognized", err_cm.exception.message)
+
 
 class TestIgblastInternal(TestBase):
     """Test igblast with internal db."""