Add files via upload

Author: omidbazgirTTU

Toolbox.py

@@ -1,7 +1,68 @@
+# -*- coding: utf-8 -*-
+"""
+Created on Mon Oct  7 18:41:14 2019
+
+@author: obazgir
+"""
+
 import numpy as np
 import pandas as pd
 import os
+import matplotlib.pyplot as plt
+from sklearn.model_selection import train_test_split
+from fancyimpute import KNN
 from scipy.stats import pearsonr
+from sklearn.linear_model import LinearRegression
+
+
+
+
+#%% NRMSE
+def NRMSE(Y_Target, Y_Predict):
+    Y_Target = np.array(Y_Target); Y_Predict = np.array(Y_Predict);
+    Y_Target = Y_Target.reshape(len(Y_Target),1);    Y_Predict = Y_Predict.reshape(len(Y_Predict),1);    
+    Y_Bar = np.mean(Y_Target)
+    Nom = np.sum((Y_Predict - Y_Target)**2);    Denom = np.sum((Y_Bar - Y_Target)**2)
+    MSE = np.mean((Y_Predict - Y_Target)**2);   NRMSE = np.sqrt(Nom/Denom)
+    R2 = 1 - NRMSE**2
+    return NRMSE, R2
+
+def NMAE(Y_Target, Y_Predict):
+    Y_Target = np.array(Y_Target); Y_Predict = np.array(Y_Predict);
+    Y_Target = Y_Target.reshape(len(Y_Target),1);    Y_Predict = Y_Predict.reshape(len(Y_Predict),1);    
+    Y_Bar = np.mean(Y_Target)
+    Nom = np.abs((Y_Predict - Y_Target));    Denom = np.abs((Y_Bar - Y_Target))
+    NormMAE = np.mean(Nom)/np.mean(Denom)
+    return NormMAE
+    
+
+#%% Random position generation
+import math
+def Random_position(p):
+    NN = int(math.sqrt(p)) +1
+    Feat_num = np.arange(p)
+    np.random.shuffle(Feat_num)
+    #Feature_List = [f'F{i}' for i in Feat_num]
+    Pos_mat = []
+    for i in range(p):
+        Pos_mat.append([int(Feat_num[i]/NN),int(Feat_num[i]%NN)])
+    return(Pos_mat)
+    
+def Random_Image_Gen(X, Rand_Pos_mat):
+    sz = X.shape
+    p = sz[1];  N = sz[0]
+    NN = int(math.sqrt(p)) +1
+    Im = np.zeros((NN,NN))
+    X_Gen = np.zeros((N,NN**2))
+    for j in range(N):
+        for i in range(p):
+            P = Rand_Pos_mat[i]
+            Im[P[0],P[1]] = X[j,i]
+        Image_Store = Im.reshape((NN**2,1)).T
+        X_Gen[j,:] = Image_Store
+    return X_Gen
+
+#%% MDS by Ruibo
 from sklearn.manifold import MDS
 
 def two_d_norm(xy):
@@ -101,3 +162,101 @@ def Assign_features_to_pixels(xy,nn,verbose = False):
 #eq_xy = two_d_eq(mds_xy)
 #Img = Assign_features_to_pixels(eq_xy,nn,verbose=1)
 #Init_Corr_MDS = InitCorr(dist_mat,Img,nn)
+
+def MDS_Im_Gen(X,nn, Img):
+    [N_sam,P_Feat] = X.shape
+    X_Gen = np.zeros((N_sam,nn**2))
+    conv_Img = Img.reshape(Img.size,1)
+    for i in range(nn**2):
+        Feature = np.array(conv_Img[i]);    Feature = Feature[0]; F_Num = int(Feature[1:])
+        if abs(F_Num) < nn**2:
+            X_Gen[:,i] = X[:,F_Num]
+        else:
+            X_Gen[:,i] = 0 
+    return X_Gen
+
+#%% CCLE functions
+def dataframer(Main,Set_in, name_in, name_out):
+    A = Set_in[name_in].tolist()
+    Set_out = Main[Main[name_out] == A[0]]
+    for cell in range(len(A) - 1):
+        df = Main[Main[name_out] == A[cell + 1]]
+        Set_out = pd.concat([Set_out, df])
+    return Set_out
+
+def Reg_to_Class(Y,Threshold):
+	Y_Class = np.zeros(len(Y))
+	Y_Sens = np.where(Y > Threshold)
+	Y_Class[Y_Sens] = 1
+	Y_Class =  Y_Class.astype(int)
+	Y_Class = Y_Class.tolist()
+	Y_Class = np.array(Y_Class)
+	return Y_Class
+	
+def floattoint(Y_Test_Encoded):
+	Y_Class = np.zeros(Y_Test_Encoded.shape)
+	Y_Sens = np.where(Y_Test_Encoded > 0.5)
+	Y_Class[Y_Sens] = 1
+	Y_Class =  Y_Class.astype(int)
+	Y_Class = Y_Class.tolist()
+	Y_Class = np.array(Y_Class)
+	return Y_Class
+def REFINED_Im_Gen(X,nn, map_in_int, gene_names,coords):
+	[N_sam,P_Feat] = X.shape
+	X_Gen = np.zeros((N_sam,nn**2))
+	for i in range(N_sam):
+		data = X[i,:] 
+		X_REFINED = pd.DataFrame(data = data.reshape(1,len(data)), columns = gene_names)
+		Image = np.zeros(map_in_int.shape)
+		for j in range(len(coords)):
+			val = np.array(X_REFINED[gene_names[j]])
+			Image[coords[j,0],coords[j,1]] = val
+		Image = Image.reshape(nn**2)
+		X_Gen[i,:] = Image
+	return X_Gen
+#%% GDSC
+def GDSC_dataframer(PD_Set, Set_Name,PD_Attribute,Attribute_Name):
+    A = PD_Set[Set_Name].tolist()
+    b = PD_Attribute[PD_Attribute[Attribute_Name] == A[0]].reset_index().drop(columns = ['index'])
+    Data_arry = np.array(b.values[0,1:],dtype = float)
+    Data_arry = Data_arry.reshape(1,len(Data_arry))
+    for i in range(len(A) - 1 ):
+        b = PD_Attribute[PD_Attribute[Attribute_Name] == A[i + 1]].reset_index().drop(columns = ['index'])
+        Arr = np.array(b.values[0,1:],dtype = float)
+        Arr = Arr.reshape(1,len(Arr))
+        Data_arry = np.append(Data_arry,Arr, axis = 0)
+    
+    PD_Data_arry = pd.DataFrame(data = Data_arry, columns = PD_Attribute.columns.tolist()[1:], index = A)
+    return Data_arry, PD_Data_arry
+	
+def GDSC_NPier(PD_Set, Set_Name,PD_Attribute,Attribute_Name):
+    PD_Set = PD_Set.reset_index()
+    PD_Set.shape[0]
+    PD_Attribute.shape[1] - 1
+    X_NP = np.zeros((PD_Set.shape[0],PD_Attribute.shape[1] - 1))
+    Source = list(set(PD_Set[Set_Name].tolist()))
+    for name in Source:
+        idx = PD_Set.index[PD_Set[Set_Name] == name].tolist()
+        XX = np.array(PD_Attribute[PD_Attribute[Attribute_Name] == name].values[0,1:], dtype = float)
+        X_NP[idx,:] = XX
+    return X_NP
+	
+	
+def Coord_Converter(coords_drug2,nn):
+    coords_drug3 = np.full((nn,nn),'NaN').astype(object)
+    for i in range(nn):
+        for j in range(nn):
+            ft = 'F' + str(coords_drug2[i,j])
+            coords_drug3[i,j] = ft
+    return coords_drug3
+	
+def Bias_Calc(Y_Test, Y_Pred):
+    Error = Y_Test - Y_Pred
+    Y_Test = Y_Test.reshape(len(Y_Test),1)
+    Error = Error.reshape(len(Error),1)
+    
+    reg = LinearRegression().fit(Y_Test, Error)
+    Bias = reg.coef_[0]
+    
+    return Bias
+