fit_line() can output LI and HI (#1)

src/pCRscore/pipeline.py

@@ -30,7 +30,7 @@ def combine_fractions_shap(data_norm, shap):
 
 
 # Function to fit a line to SHAP vs Fraction for each cell type
-def fit_line(data):
+def fit_line(data, split_ci=False):
     result = []
     grouped = data.groupby('Feature')
     for name, group in grouped:
@@ -39,7 +39,14 @@ def fit_line(data):
         model = statsmodels.api.OLS(y, X).fit()
         coef = model.params['Fraction']
         ci = model.conf_int(alpha=0.001).loc['Fraction']
-        result.append({'Feature': name, 'Coef': coef, 'CI': ci[0] * ci[1]})
+        if split_ci:
+            result.append(
+                {'Feature': name, 'Coef': coef, 'LI': ci[0], 'HI': ci[1]}
+            )
+        else:
+            result.append(
+               {'Feature': name, 'Coef': coef, 'CI': ci[0] * ci[1]}
+            )
     return pandas.DataFrame(result)
 
 

tests/test_pipeline.py

@@ -133,3 +133,26 @@ def test_combine_discovery_validation():
     assert 'SHAP value' in all_pat.columns
     if local:
         all_pat.shape == (15888, 3)
+
+
+# A new fitting is performed on discovery and validation data
+fit = pipeline.fit_line(all_pat, split_ci=True)
+
+
+def test_fit():
+    assert fit.shape[0] == len(all_pat['Feature'].unique())
+    assert 'Feature' in fit.columns
+    assert 'Coef' in fit.columns
+    assert 'LI' in fit.columns
+    assert 'HI' in fit.columns
+    if local:
+        line = fit['Feature'] == 'CAFs.myCAF.like'
+        assert math.isclose(
+            fit.loc[line, 'Coef'].values[0], -0.028020, rel_tol=1e-2
+        )
+        assert math.isclose(
+            fit.loc[line, 'LI'].values[0], -0.041878, rel_tol=1e-2
+        )
+        assert math.isclose(
+            fit.loc[line, 'HI'].values[0], -0.014161, rel_tol=1e-2
+        )