adjust sample filtering parameter

jass_preprocessing/__main__.py

@@ -83,7 +83,7 @@ def add_preprocessing_argument():
 
     parser.add_argument('--output-folder', required=True, help= "Location of main ouput folder for preprocessed GWAS files (splitted by chromosome)")
     parser.add_argument('--output-folder-1-file', required=False, help= "optional location to store the preprocessing in one tabular file with one chromosome columns (useful to compute LDSC correlation for instance)")
-    parser.add_argument('--percent-sample-size', required=False, help= "the proportion (between 0 and 1) of the 90th percentile of the sample size used to filter the SNPs", default=0.7)
+    parser.add_argument('--percent-sample-size', required=False, help= "the proportion (between 0 and 1) of the 90th percentile of the sample size used to filter the SNPs", default=0.75)
 
     parser.add_argument('--minimum-MAF', required=False, help= "Filter the reference panel by  minimum allele frequency", default='0.01')
     parser.add_argument('--mask-MHC', required=False, help= "Whether the MHC region should be masked or not. default is False", default='False')

jass_preprocessing/compute_score.py

@@ -27,8 +27,11 @@ def compute_z_score(mgwas):
 
     return mgwas
 
-def compute_sample_size(mgwas, diagnostic_folder, trait, max_sample_size_ratio = 0.1):
-
+def compute_sample_size(mgwas, diagnostic_folder, trait, min_sample_size_ratio = 0.95):
+    """
+    Infer sample size when not present and filter sample size to reach homogeneous sample
+    size
+    """
     if 'n' in mgwas.columns:
         myN = mgwas.n
     #--- freq, case-cont N exist
@@ -58,13 +61,15 @@ def compute_sample_size(mgwas, diagnostic_folder, trait, max_sample_size_ratio =
     myW_thres_low = np.percentile(myN.dropna(), perc_low)
     myW_thres_max = np.percentile(myN.dropna(), perc_max)
 
-    while (1 - (myW_thres_low /myW_thres_max)) >  max_sample_size_ratio: # narrow treshold until sample size can be considered as homogeneous
-        perc_low += 10
-        perc_max -= 10
+    while ((myW_thres_low / myW_thres_max) < min_sample_size_ratio): # narrow treshold until sample size can be considered as homogeneous
+        perc_low += 2
+        perc_max -= 0.5
 
         myW_thres_low = np.percentile(myN.dropna(), perc_low)
         myW_thres_max = np.percentile(myN.dropna(), perc_max)
 
+
+    print("Finished filtering")
     mgwas["computed_N"] = myN
     plt.clf()
     p1 = sns.distplot(mgwas.computed_N[~mgwas.computed_N.isna()])

jass_preprocessing/save_output.py

@@ -19,8 +19,9 @@ def save_output_by_chromosome(mgwas, ImpG_output_Folder, my_study):
                             'A0': mgwas_copy.loc[chrom].ref,
                             'A1':mgwas_copy.loc[chrom].alt,
                             'Z': mgwas_copy.loc[chrom].computed_z,
-                            'P': mgwas_copy.loc[chrom].pval
-                }, columns= ['rsID', 'pos', 'A0', "A1", "Z", "P" ])
+                            'P': mgwas_copy.loc[chrom].pval,
+                            'N_effective': mgwas_copy.loc[chrom].computed_N
+                }, columns= ['rsID', 'pos', 'A0', "A1", "Z", "P", "N_effective"])
 
             impg_output_file = ImpG_output_Folder + 'z_'+ my_study +'_chr'+str(chrom)+".txt"
             print("WRITING CHR {} results for {} to: {}".format(chrom, my_study, ImpG_output_Folder))
@@ -43,8 +44,9 @@ def save_output(mgwas, ImpG_output_Folder, my_study):
                     'A0': mgwas_copy.ref,
                     'A1':mgwas_copy.alt,
                     'Z': mgwas_copy.computed_z,
-                    'P': mgwas_copy.pval
-        }, columns= ['chrom','rsID', 'pos', 'A0', "A1", "Z", "P" ])
+                    'P': mgwas_copy.pval,
+                    'N_effective': mgwas_copy.computed_N
+        }, columns= ['chrom','rsID', 'pos', 'A0', "A1", "Z", "P", 'N_effective' ])
 
     impg_output_file = ImpG_output_Folder + 'z_'+ my_study +".txt"
     print("WRITING results for {} to: {}".format( my_study, ImpG_output_Folder))