关于EeepOI指标及CCL指标验证

Posted 2022-06-15 Mario cai

将fprint进行排序

n=1
data_fprint = np.load(fr'D:\\DEPOI\\npy_data\\n\\fprintn.npy', allow_pickle=True)
x,y = data_fprint.shape
# print((data_fprint[5] == data_fprint[88]).all())

print(x,y)
count=[]
z=-1
while z <= 1297 :
    z = z+1
    for i in range(z, x):
        if((data_fprint[z] == data_fprint[i]).all()):

            count.append(i)
    print(z)
record = pd.DataFrame(count)  # predictive   real
record.to_csv(fr'D:\\DEPOI\\npy_data\\n\\fprintn.csv', index=False)

当模型输出结果之后，寻找fprint与sample进行匹配。

例子是用预测值和fprint匹配。

# n =2
# df = pd.read_csv(fr'D:\\DEPOI\\npy_data\\n\\fprintn.csv',encoding='gbk')
# test = pd.read_csv(fr'D:\\DEPOI\\npy_data\\n\\crossn_pred.csv',encoding='gbk') 
#
# df = np.array(df)
# test = np.array(test)
#
# test = np.transpose(test,(1,0))
# # df = np.transpose(df,(1,0))
#
#
# # print(test[0][df[1][1]])
# x,y = df.shape
# print(df[0])
# print(x,y)
# count = []
# T=[]
# print(test[0][1297])
#
#
# for j in range(0,x):
#     # for i in range(0, y):
#     # print(test[0][df[x][i]])
#     # t = test[0][df[x][i]]
#     # print(df[x][i])
#     # print(df[j])
#     t = np.array(test[0][df[j]])
#     # print(i,j)
#
#     # print(t)
#     # T.append(t)
#     # break
#
#
#     count.append(t)
#
# # #
# count = np.array(count)
# record = pd.DataFrame(count)  # predictive   real
# record.to_csv(fr'D:\\DEPOI\\npy_data\\n\\n_pred.csv', index=False)

匹配完成之后将模型输出的结果切割成基因形状Deepoi=1298

n=2
df = pd.read_csv(fr'D:\\DEPOI\\npy_data\\n\\n_pred.csv',encoding='gbk')

T = np.array(df.real)
T = np.array_split(T,1298)
# T = sum(T)
record = pd.DataFrame(T)  # predictive   real
record.to_csv(fr'D:\\DEPOI\\npy_data\\n\\ttt.csv', index=False)

P = np.array(df.pred)
P = np.array_split(P,1298)
# P = sum(P)
record1 = pd.DataFrame(P)  # predictive   real
record1.to_csv(fr'D:\\DEPOI\\npy_data\\n\\ppp.csv', index=False)

r,p = stats.pearsonr(T,P)## 相关系数和P值
print('相关系数r为 = %6.3f，p值为 = %6.3f'%(r,p))

输出为两个表格之后，利用np.transpose进行翻转测试。

n=2
P = pd.read_csv(fr'D:\\DEPOI\\npy_data\\n\\ppp.csv')
T = pd.read_csv(fr'D:\\DEPOI\\npy_data\\n\\ttt.csv')


P = np.array(P)
T = np.array(T)
# P = np.transpose(P,(1,0))
# T = np.transpose(T,(1,0))

x,y = P.shape
print(x,y)
count = []
t = []
p = []
# print(T[1])
for j in range(0,x):
    r,pp = stats.pearsonr(T[j],P[j])## 相关系数和P值
    count.append(r)
#
count = np.array(count)
record = pd.DataFrame(count)  # predictive   real
record.to_csv(fr'D:\\DEPOI\\npy_data\\n\\deepoi4.csv', index=False)