selection colomn invariant prosit data

alignement.py

@@ -45,6 +45,16 @@ def numerical_to_alphabetical(arr):
         seq+=ALPHABET_UNMOD_REV[arr[i]]
     return seq
 
+def numerical_to_alphabetical_str(s):
+    seq = ''
+    s = s.replace('[','')
+    s = s.replace(']', '')
+    arr = s.split(',')
+    arr = list(map(int, arr))
+    for i in range(len(arr)):
+        seq+=ALPHABET_UNMOD_REV[arr[i]]
+    return seq
+
 def align(dataset, reference):
     seq_ref = reference['Sequence']
     seq_common = dataset['Sequence']
@@ -87,7 +97,7 @@ def compare_include_df(df, sub_df, save = True, path = 'temp.png'):
             df_sub_value_list.append(r[1]['Retention time'])
         except:
             pass
-
+    plt.clf()
     fig, ax = plt.subplots()
     ax.scatter(df_sub_value_list, df_value_list)
     x = np.array([min(df_value_list), max(df_value_list)])
@@ -183,16 +193,44 @@ def compare_include_df(df, sub_df, save = True, path = 'temp.png'):
 # data_align.to_pickle('database/data_ISA_dual_align.pkl')
 
 
+#compare DIANN pred to DIA mesures
+# df_ori = pd.read_csv('database/data_train.csv')
+# df_ori['Sequence']=df_ori['sequence']
+# df_ori['Retention time']=df_ori['irt']
+# df_diann = pd.read_csv('database/CIT_BASE_UP000584719_546.csv')
+#
+# df_ISA = pd.read_pickle('database/data_ISA_dual_align.pkl')
+#
+#
+#
+# df_diann_aligned = align(df_diann, df_ori)
+#
+# df_value_list, df_sub_value_list = compare_include_df(df_diann_aligned, df_ISA, True)
+
+
+
+
+#create augmented dataset from ISA data + column invariant prosit peptides
+df_base = pd.read_pickle('database/data_ISA_aligned_prosit.pkl')
+df_base = df_base[['Sequence','Retention time']]
 
-df_ori = pd.read_csv('database/data_train.csv')
-df_ori['Sequence']=df_ori['sequence']
-df_ori['Retention time']=df_ori['irt']
-df_diann = pd.read_csv('database/CIT_BASE_UP000584719_546.csv')
+df_1 = pd.read_pickle('database/data_prosit_threshold_1.pkl')
+df_1['Sequence']= df_1['Sequence'].map(numerical_to_alphabetical_str)
 
-df_ISA = pd.read_pickle('database/data_ISA_dual_align.pkl')
+df_2 = pd.read_pickle('database/data_prosit_threshold_2.pkl')
+df_2['Sequence']= df_2['Sequence'].map(numerical_to_alphabetical_str)
 
+df_3 = pd.read_pickle('database/data_prosit_threshold_3.pkl')
+df_3['Sequence']= df_3['Sequence'].map(numerical_to_alphabetical_str)
 
+df_augmented_1 = pd.concat([df_1,df_base],axis=0).reset_index(drop=True)
+df_augmented_1.columns=['sequence','retention_time']
+df_augmented_1.to_csv('database/data_ISA_augmented_1.csv')
 
-df_diann_aligned = align(df_diann, df_ori)
+df_augmented_2 = pd.concat([df_2,df_base],axis=0).reset_index(drop=True)
+df_augmented_2.columns=['sequence','retention_time']
+df_augmented_2.to_csv('database/data_ISA_augmented_2.csv')
 
-df_value_list, df_sub_value_list = compare_include_df(df_diann_aligned, df_ISA, True)
+df_augmented_3 = pd.concat([df_3,df_base],axis=0).reset_index(drop=True)
+df_augmented_3.columns=['sequence','retention_time']
+df_augmented_3.to_csv('database/data_ISA_augmented_3.csv')
\ No newline at end of file

data_exploration.py

@@ -287,11 +287,31 @@ def RT_distrib(Y, f_name):
 #
 #ISA DATA
 
-df = pd.read_pickle('database/data_ISA_aligned_prosit.pkl')
-seq = df['Sequence'].unique()
+# df = pd.read_pickle('database/data_ISA_aligned_prosit.pkl')
+# seq = df['Sequence'].unique()
 # rt = df['Retention time']
-df_mean = df.groupby(['Sequence'])['Retention time'].mean()
+# df_mean = df.groupby(['Sequence'])['Retention time'].mean()
 # feq_aa(seq, plot=False, save=True, f_name='fig/histo_aa_ISA_unique.png')
 # dist_long(seq, plot=False, save=True, f_name='fig/histo_length_ISA_unique.png')
 
-RT_distrib(df_mean, 'fig/histo_RT_ISA_unique.png')
+# RT_distrib(df_mean, 'fig/histo_RT_ISA_unique.png')
+
+
+df_ISA = pd.read_pickle('database/data_ISA_dual_align.pkl')
+seq_list = []
+var_list = []
+num_list = []
+len_list=[]
+for seq, gr in df_ISA.groupby(['Sequence']) :
+    if len(gr)>10 :
+        seq_list.append(seq)
+        var_list.append(np.std(gr['Retention time']))
+        num_list.append(len(gr))
+        len_list.append(len(seq))
+idx = np.argsort(var_list)
+seq_list = np.array(seq_list)[idx][::-1]
+var_list = np.array(var_list)[idx][::-1]
+num_list = np.array(num_list)[idx][::-1]
+len_list = np.array(len_list)[idx][::-1]
+
+

data_viz.py

@@ -252,11 +252,9 @@ def select_best_data(df1,df2,threshold):
     df = pd.concat([df_group_1, df_group_2], axis=1)
     df['mean']=(df['abs err 1']+df['abs err 2'])/2
     df_res = df[df['mean']<threshold]
-    print(df_res.size)
     df_res = df_res['seq']
     df_res.columns = ['seq','temp']
     df_res = df_res['seq']
-
     good_seq=[]
     good_rt=[]
     for r in df1.iterrows() :
@@ -312,7 +310,4 @@ df_2 = pd.read_csv('output/out_common_ISA_prosit_eval_2.csv')
 
 df = select_best_data(df_1, df_2, 3)
 df.to_pickle('database/data_prosit_threshold_3.pkl')
-# compare_error(df_1,df_2,save=True,path='fig/custom model res/ISA_prosit_error_variation.png')
-
-
-
+# compare_error(df_1,df_2,save=True,path='fig/custom model res/ISA_prosit_error_variation.png')
\ No newline at end of file

seq_cutting.py

@@ -57,4 +57,4 @@ def digest(seq, format):
     ind = cut(seq, format)
     return cut_with_ind(seq, ind)
 
-res = digest('MNPLLILTFVAAALAAPFDDDDKIVGGYNCEENSVPYQVSLNSGYHFCGGSLINEQWVVSAGHCYKSRIQVRLGEHNIEVLEGNEQFINAAKIIRHPQYDRKTLNNDIMLIKLSSRAVINARVSTISLPTAPPATGTKCLISGWGNTASSGADYPDELQCLDAPVLSQAKCEASYPGKITSNMFCVGFLEGGKDSCQGDSGGPVVCNGQLQGVVSWGDGCAQKNKPGVYTKVYNYVKWIKNTIAANS','alphabetical')
\ No newline at end of file
+res = digest('FPTDDDDKIVGGYTCAANSIPYQVSLNSGSHFCGGSLINSQWVVSAAHCYKSRIQVRLGEHNIDVLEGNEQFINAAKIITHPNFNGNTLDNDIMLIKLSSPATLNSRVATVSLPRSCAAAGTECLISGWGNTKSSGSSYPSLLQCLKAPVLSDSSCKSSYPGQITGNMICVGFLEGGKDSCQGDSGGPVVCNGQLQGIVSWGYGCAQKNKPGVYTKVCNYVNWIQQTIAA','alphabetical')
\ No newline at end of file