replace inf computed_Z score by real Z score or maximum value

jass_preprocessing/compute_score.py

@@ -12,18 +12,36 @@ def compute_z_score(mgwas):
     """
     Compute zscore value and sign1
     add the corresponding column to the mgwas dataframe
+    
+    the smallest positive value of float
+    sys.float_info.min
+    2.2250738585072014e-308
+
+    the biggest Z score
+    np.sqrt(ss.chi2.isf(sys.float_info.min, 1))
+    37.537836095576054
     """
+    
+    print(mgwas.columns)
+    mgwas["computed_z"] = np.sqrt(ss.chi2.isf(mgwas['pval'], 1))
+
+    if 'z' in mgwas.columns:
+        sign_vect = np.sign(mgwas.z)
+        
+        mgwas.loc[mgwas["computed_z"].isin([np.inf]), 'computed_z'] = mgwas["z"]
 
-    if 'beta_or_Z' in mgwas.columns:
-        sign_vect = np.sign(mgwas.beta_or_Z)
+    elif 'beta' in mgwas.columns:
+        sign_vect = np.sign(mgwas.beta)
+        mgwas["computed_z"] = mgwas["computed_z"].replace(np.inf, 37.537836095576054)
     else:
         if "OR" in mgwas.columns:
             sign_vect = np.sign(mgwas["OR"] - 1.0 + 10**(-8))
+            mgwas["computed_z"] = mgwas["computed_z"].replace(np.inf, 37.537836095576054)
         else:
             raise ValueError(
                 'The gwas data frame doesn"t contain effect column')
 
-    mgwas["computed_z"] = np.sqrt(ss.chi2.isf(mgwas['pval'], 1)) * sign_vect * mgwas["sign_flip"]
+    mgwas["computed_z"] = mgwas["computed_z"] * sign_vect * mgwas["sign_flip"]
 
     return mgwas