If the presentation models are downloaded but not the affinity models, use the...

If the presentation models are downloaded but not the affinity models, use the affinity predictor included with the presentation models. Also, bump version to 1.6.1.

mhcflurry/class1_affinity_predictor.py

@@ -453,7 +453,19 @@ class Class1AffinityPredictor(object):
         `Class1AffinityPredictor` instance
         """
         if models_dir is None:
-            models_dir = get_default_class1_models_dir()
+            try:
+                models_dir = get_default_class1_models_dir()
+            except RuntimeError as e:
+                # Fall back to the affinity predictor included in presentation
+                # predictor if possible.
+                from mhcflurry.class1_presentation_predictor import (
+                    Class1PresentationPredictor)
+                try:
+                    presentation_predictor = Class1PresentationPredictor.load()
+                    return presentation_predictor.affinity_predictor
+                except RuntimeError:
+                    raise e
+
         if optimization_level is None:
             optimization_level = OPTIMIZATION_LEVEL
 

mhcflurry/version.py

-__version__ = "1.6.0"
+__version__ = "1.6.1"