python_93040206/admin_site/service/utils.py

import pandas as pd
from pandas.api.types import is_string_dtype
from pandas.api.types import is_numeric_dtype
from sklearn.preprocessing import LabelEncoder
import matplotlib.pyplot as plt
from sklearn.preprocessing import MinMaxScaler
from sklearn.model_selection import train_test_split
import os
import pyreadr
from sklearn.linear_model import LogisticRegression
import joblib
from sklearn.metrics import roc_curve, auc
import numpy as np


def ROC_curve(clf, X_test, y_test, name):
    # ROC曲线绘制,AUC值计算
    # 计算每一类的得分
    yscore_rf = clf.predict_proba(X_test)
    fpr_rf, tpr_rf, thersholds = roc_curve(y_test, yscore_rf[:, 1])
    plt.figure()
    plt.plot(fpr_rf, tpr_rf, color='darkorange',
             lw=2, label='ROC curve (area = %0.2f)' % auc(fpr_rf, tpr_rf))
    plt.plot([0, 1], [0, 1], color='navy', lw=2, linestyle='--')
    plt.xlim([0.0, 1.0])
    plt.ylim([0.0, 1.05])
    plt.xlabel('False Positive Rate')
    plt.ylabel('True Positive Rate')
    # plt.title('Receiver operating characteristic example')
    plt.title(name)
    plt.legend(loc="lower right")
    # plt.savefig(plt.savefig('figs/savefig_example.png'))
    plt.savefig('../images/' + name + '.png')
    # plt.show()


def save_gene_list(COAD_path, STAD_path):
    print(COAD_path, STAD_path)
    # 读取非胃癌数据
    data_health = pyreadr.read_r(COAD_path)
    data_health = data_health['data']  # .head(10000) # 提取前10000行，后续把.head()删掉
    # 读取胃癌数据
    data_sick = pyreadr.read_r(STAD_path)
    data_sick = data_sick['data']  # .head(10000) # 提取前10000行，后续把.head()删掉

    # 数据处理
    data_health['D'] = 0
    data_sick['D'] = 1
    data = pd.concat([data_health, data_sick])
    print(50, type(data))
    data.to_csv('data.csv')
    # data_sick[['Variant_Classification','Tumor_Sample_Barcode']]
    sample_list = list(data.groupby(['Tumor_Sample_Barcode']).groups.keys())  # 把获取类别转为列表
    gene_list = list(data.groupby(['Variant_Classification']).groups.keys())  # 把获取类别转为列表
    np.save('sample_list.npy', np.array(sample_list))  # 保存为.npy格式
    np.save('gene_list.npy', np.array(gene_list))  # 保存为.npy格式

def get_gene_list():
    gene_list = np.load('gene_list.npy')
    gene_list = gene_list.tolist()
    return gene_list
def get_sample_list():
    sample_list = np.load('sample_list.npy')
    sample_list = sample_list.tolist()
    return sample_list

def get_data():
    snp_data = pd.read_csv('data.csv')  # 假设SNP数据已保存在CSV文件中
# save_gene_list(COAD_path='../Rdata/COAD.Rdata', STAD_path='../Rdata/STAD.Rdata')

# 计算基因数目并保存
def count_the_number_of_genes():
    data = pd.read_csv('data.csv', low_memory=False)  # 假设SNP数据已保存在CSV文件中
    gene_list = get_gene_list()
    sample_list = get_sample_list()
    list_final = gene_list + ['Tumor_Sample_Barcode', 'D']
    data_final = pd.DataFrame(columns=list_final)
    ind = 0
    for sample in sample_list:
        tempdata = data[data['Tumor_Sample_Barcode'] == sample]
        message = []
        for gene in gene_list:
            num = len(tempdata[tempdata['Variant_Classification'] == gene])
            message.append(num)
        message = message + [sample, tempdata.iloc[0][-1]]
        data_final.loc[ind] = message
        ind = ind + 1
    data_final.to_csv('data_final.csv')
# count_the_number_of_genes()
# 数据处理
def load_data(COAD_path, STAD_path):
    # 读取非胃癌数据
    data_health = pyreadr.read_r('COAD.Rdata')
    data_health = data_health['data']  # .head(10000) # 提取前10000行，后续把.head()删掉
    # 读取胃癌数据
    data_sick = pyreadr.read_r('STAD.Rdata')
    data_sick = data_sick['data']  # .head(10000) # 提取前10000行，后续把.head()删掉


def training_SC_GRS_model():
    data_final = pd.read_csv('data_final.csv', low_memory=False)  # 假设SNP数据已保存在CSV文件中
    gene_list = get_gene_list()
    data_final['Gi_sum'] = data_final[gene_list].sum(axis=1)
    # 模型训练
    X_train, X_test, y_train, y_test = train_test_split(data_final['Gi_sum'].values.reshape(-1, 1),
                                                        data_final['D'].values.astype('int'), test_size=0.2)
    clf = LogisticRegression(penalty='l2')
    clf.fit(X_train, y_train)
    # 模型系数
    coef = clf.coef_
    intercept = clf.intercept_
    # 储存模型
    joblib.dump(clf, 'clf_scgrs.model')

    # ROC绘图 SC_GRS方法
    ROC_curve(clf, X_test, y_test, name='SC_GRS')
    print('training_SC_GRS_model')

    # 模型预测,假设data_final是输入数据，注意，data_final已经是处理好对应的基因数目的数据
    # 模型加载
    gene_list = np.load('gene_list.npy')
    gene_list = gene_list.tolist()
    clf_scgrs = joblib.load('clf_scgrs.model')
    # 数据处理
    X_test = data_final[gene_list].sum(axis=1).values.reshape(-1, 1)
    y_pre = clf.predict(X_test)  # 分类结果预测
    prob_pre = clf.predict_proba(X_test)  # 结果概率预测
    # print(95, prob_pre)
    # print(95, np.amin(prob_pre))
    # print(95, np.amax(prob_pre))
    print(95, np.mean(prob_pre))
    print(95, np.mean(y_pre))

training_SC_GRS_model()

def training_DL_GRS_model():
    print('training_DL_GRS_model')


def training_OR_GRS_model():
    print('training_OR_GRS_model')