基于序列进化信息的细菌毒力因子预测研究

# 核心功能实现代码
（1）PSI-blast批量生成
names=[name for name in os.listdir('//home/bhliu/data/negativetraindata1/01//') if os.path.isfile(os.path.join('//home/bhliu/data/negativetraindata1/01//', name))]
for each_item in names:
    uniprotid=each_item.split('.')[0]
    cmd='/home/bhliu/ncbi-blast-2.12.0+/bin/psiblast  -num_iterations 3 -db /home/bhliu/uniref50.fasta -query  /home/bhliu/data/negativetraindata1/01/'+uniprotid+'.fasta -out_ascii_pssm /home/bhliu/pssm/negativepssm/'+uniprotid+'.pssm '
    #print(cmd)
    os.system(cmd)
（2）特征提取
def PSSM_composition(proteinSeq_pro,proteinSeq,pssmMatrix):
	#pssm_composition特征提取
	PSSM_composition=[]
	pssm_composition=[[0.0for m in range(20)] for n in range(20)]#创建一个20*20且元素都是0.0的矩阵
	for k in range(20):
		for i in range(len(proteinSeq)):
			if proteinSeq_pro[k]==proteinSeq[i]:
				for j in range(20):			pssm_composition[k][j]=pssm_composition[k][j]+pssmMatrix[i][j]/len(pssmMatrix)
	for a in pssm_composition:
		for b in a:
			PSSM_composition.append(b)
	return PSSM_composition

def S_FPSSM(proteinSeq_pro,proteinSeq,pssmMatrix):
	#S_FPSSM特征提取
	F_PSSM=pssmMatrix#转换F_PSSM矩阵
	for k in range(20):
		for i in range(len(proteinSeq)):
			if F_PSSM[i][k]<=0.0:
				F_PSSM[i][k]=0.0
			elif F_PSSM[i][k]>=7.0:
				F_PSSM[i][k] = 7.0
	S_fpssm=[]
	s_fpssm=[[0.0for m in range(20)] for n in range(20)]
	for k in range(20):
		for i in range(len(proteinSeq)):
			if proteinSeq_pro[k]==proteinSeq[i]:
				for j in range(20):
					s_fpssm[k][j]=s_fpssm[k][j]+F_PSSM[i][j]
	for a in s_fpssm:
		for b in a:
			S_fpssm.append(b)
	return S_fpssm

def RPM_PSSM(proteinSeq_pro,proteinSeq,pssmMatrix):
	#RPM-PSSM特征提取
	PPSSM=pssmMatrix#将原始PSSM矩阵转换为PPSSM
	RPM_PSSM=[]
	rpm_PSSM=[[0.0for m in range(20)] for n in range(20)]
	for k in range(20):
		for i in range(len(proteinSeq)):
			if PPSSM[i][k]<=0.0:
				PPSSM[i][k]=0
	for k in range(20):
		for i in range(len(proteinSeq)):
			if proteinSeq_pro[k]==proteinSeq[i]:
				for j in range(20):
					rpm_PSSM[k][j]=rpm_PSSM[k][j]+PPSSM[i][j]/len(proteinSeq)
	for a in rpm_PSSM:
		for b in a:
			RPM_PSSM.append(b)
	return RPM_PSSM
（3）打分函数
def model_process(svm_model, test_data_x, test_data_y,cv_num,model_name):
    p_lable = svm_model.predict(test_data_x)
    print('总体精度为 : {}'.format(cv_num))
    print('混淆矩阵为 :\n {}'.format(confusion_matrix(test_data_y, p_lable)))
    print('kappa系数为 :\n {}'.format(cohen_kappa_score(test_data_y, p_lable)))
    matric = confusion_matrix(test_data_y, p_lable)
    TP = matric[1, 1] + 0.0
    TN = matric[0, 0] + 0.0
    FP = matric[0, 1] + 0.0
    FN = matric[1, 0] + 0.0
    recall = TP / (TP + FN)
    precise = TP / (TP + FP)
    specificity = TN / (TN + FP)
    f1_score = 2 * (TP / (2 * TP + FP + FN))
    MCC = ((TP * TN) - (FN * FP)) / math.sqrt((TP + FN) * (TN + FP) * (TP + FP) * (TN + FN))
    print('{}的召回率（recall）为{:.4}，特异度（specificity）为{:.4}，精确率（precision）为{:.4}，F1 score为{:.4}，MCC值为{:.4} '.format(model_name,                                                                                                              recall,specificity,precise,f1_score,MCC))
    print(p_lable.tolist().count(0))
（4）交叉验证模型性能
def model_cv(path):
    Pssm=csv_read.csv_read(path)
    from sklearn.preprocessing import  StandardScaler
    scaler=StandardScaler()#实例化
    scaler.fit(Pssm)#fit，本质是生成均值和方差
    # scaler.mean_#查看均值的属性mean_
    # scaler.var_#查看方差的属性var_
    pssm=scaler.transform(Pssm)#通过接口导出结果
    y=[1for m in range(4000)]+[0for n in range(4000)]#1为毒力因子。0为非毒力因子
    y=np.array(y)

    train_x, test_x, train_y, test_y = train_test_split(pssm, y, test_size=0.2,random_state=420)


    if path == 's_fpssm_train.csv':
        j = 'S-FPSSM'
        svc_model