PSSM特征-从生成到处理

以下代码均为个人原创，如有疑问，欢迎交流。新浪微博：拾毅者

本节内容：

• pssm生成
• pssm简化
• 标准的pssm构建
• 滑动pssm生成

在基于蛋白质序列的相关预测中，使用PSSM打分矩阵会得将预测效果大大提高，同时，如果使用滑动的PSSM，效果又会进一步提高。这里主要以分享代码为主，下面介绍下PSSM从生成到处理的全过程。

1.PSSM的生成

PSSM的生成有多种方式，这里使用的psiblast软件，ncbi-blast-2.2.28+/bin/psiblast，下载地址：http://blast.ncbi.nlm.nih.gov/Blast.cgi?CMD=Web&PAGE_TYPE=BlastNews#1 使用方法，输入一个序列，加上相关参数，直接输出PSSM文件

代码

#一个命令函数，根据pdb文件，使用blast生成pssm文件
def command_pssm(content, output_file,pssm_file):
    os.system('/ifs/share/lib/blast/ncbi-blast-2.2.28+/bin/psiblast \
                -query %s \
                -db /ifs/data/database/blast_data/nr \
                -num_iterations 3 \
                -out %s \
                -out_ascii_pssm %s &' %(content, output_file,pssm_file))

上面是执行的命令，封装成函数，下面为实际代码：

#对每个序列进行PSSM打分
def pssm(proseq,outdir):
    inputfile = open(proseq,'r')
    content = ''
    input_file = ''
    output_file = ''
    pssm_file = ''
    chain_name = []
    for eachline in inputfile:
        if '>' in eachline:
            if len(content):
                temp_file = open(outdir + '/fasta/' + chain_name,'w')
                temp_file.write(content)
                input_file = outdir + '/fasta/' + chain_name
                output_file = outdir + '/' + chain_name + '.out'
                pssm_file = outdir + '/' + chain_name + '.pssm'                
                command_pssm(input_file, output_file,pssm_file)
                temp_file.close
            content = ''
            chain_name = eachline[1:5] + eachline[6:7]
        content +=  ''.join(eachline)
        #print content
        #print chain_name
    if len(content):
        temp_file = open(outdir + '/fasta/' + chain_name,'w')
        temp_file.write(content)
        input_file = outdir + '/fasta/' + chain_name
        output_file = outdir + '/' + chain_name + '.out'
        pssm_file = outdir + '/' + chain_name + '.pssm'
        command_pssm(input_file, output_file,pssm_file)  
        temp_file.close
    inputfile.close()       

测试用例：

'''            
#生成pssm文件，迭代次数为3       
proseq = '/ifs/home/liudiwei/experiment/step2/data/protein.seq'
outdir = '/ifs/home/liudiwei/experiment/step2/pssm'
pssm(proseq,outdir)
'''

PSSM输出样例：

2.简化PSSM数据

通常我们需要的只是前面的20列

下面通过代码来实现上面的功能：

#格式化pssm每行数据
def formateachline(eachline):
    col = eachline[0:5].strip()
    col += '\t' + eachline[5:8].strip()    
    begin = 9 
    end = begin +3
    for i in range(20):     
        begin = begin
        end = begin + 3
        col += '\t' + eachline[begin:end].strip()
        begin = end
    col += '\n'
    return col

简化pssm，只要得到前面的20个氨基酸的打分值

def simplifypssm(pssmdir,newdir):
    listfile = os.listdir(pssmdir)
    for eachfile in listfile:
        with open(pssmdir + '/' + eachfile,'r') as inputpssm:
            with open(newdir + '/' + eachfile,'w') as outfile:
                count = 0
                for eachline in inputpssm:
                    count +=1
                    if count <= 3:
                        continue
                    if not len(eachline.strip()):
                        break
                    oneline = formateachline(eachline)
                    outfile.write(''.join(oneline))
''' Test example
pssmdir = '/ifs/home/liudiwei/experiment/step2/pssm/oldpssm'
newdir = '/ifs/home/liudiwei/experiment/step2/pssm/newpssm'
simplifypssm(pssmdir, newdir)
'''

3.得到标准的PSSM

通过上面抽取出来的PSSM，下面通过代码来获得一个滑动的PSSM

#标准的pssm,直接根据标准的pssm滑动
def standardPSSM(window_size,pssmdir,outdir):
    listfile = os.listdir(pssmdir)
    for eachfile in listfile:
        outfile = open(outdir + '/' + eachfile, 'w') 
        with open(pssmdir + '/' + eachfile, 'r') as inputf:
            inputfile = inputf.readlines()
            for linenum in range(len(inputfile)):
                content = []
                first = [];second = [];third=[];last=[]
                if linenum < window_size/2:
                    for i in range((window_size/2 - linenum)*20):
                        second.append('\t0')
                if window_size/2 - linenum > 0:
                    countline = window_size - (window_size/2 - linenum)
                else:
                    countline = window_size  #get needed line count

                linetemp = 0
                for eachline in inputfile:
                    if linetemp < linenum-window_size/2:
                        linetemp += 1
                        continue
                    if linetemp == linenum:
                        thisline = eachline.split('\t')
                        for j in range(0,2):
                            if j>0:
                                first.append('\t')
                            first.append(thisline[j].strip())
                    if countline > 0:
                        oneline = eachline.split('\t')
                        for j in range(2,len(oneline)):
                            third.append('\t' + oneline[j].strip())
                        countline -=1
                    else:
                        break
                    linetemp += 1
                while countline:
                    for i in range(20):
                        last.append('\t0')
                    countline -=1
                content += first + second + third + last
                outfile.write(''.join(content) + '\n')
        outfile.close()

'''Test example
pssmdir = '/ifs/home/liudiwei/experiment/step2/pssm/newpssm'
newdir = '/ifs/home/liudiwei/experiment/step2/pssm/standardpssm'
window_size = 5
standardPSSM(window_size,pssmdir, newdir)
'''

4.根据滑动窗口求出滑动的PSSM

#根据窗口大小,计算出滑动后的20个氨基酸打分值
def computedPSSM(window_size,pssmdir,outdir):
    listfile = os.listdir(pssmdir)
    for eachfile in listfile:
        outfile = open(outdir + '/' + eachfile, 'w') 
        with open(pssmdir + '/' + eachfile, 'r') as inputf:
            inputfile = inputf.readlines()
            for linenum in range(len(inputfile)):
                content = []
                first = [];second = []
                if window_size/2 - linenum > 0:
                    countline = window_size - (window_size/2 - linenum)
                else:
                    countline = window_size  #get needed line count

                linetemp = 0
                for eachline in inputfile:
                    if linetemp < linenum-window_size/2:
                        linetemp += 1
                        continue
                    if linetemp == linenum:
                        thisline = eachline.split('\t')
                        for j in range(0,2):
                            if j>0:first.append('\t')
                            first.append(thisline[j].strip())
                    if countline > 0: 
                        oneline = eachline.split('\t')[2:len(eachline)]
                        tline = []
                        for i in range(len(oneline)):
                            tline.append(int(oneline[i]))
                        if len(second)==0:
                            second += tline
                        else:
                            second = list(map(lambda x: x[0]+x[1], zip(second, tline)))
                        countline -=1 
                    else:
                        break 
                    linetemp += 1 
                format_second = []
                for i in range(len(second)):
                    format_second.append('\t' + str(second[i]))
                content += first + format_second 
                outfile.write(''.join(content) + '\n')
        outfile.close()
'''      
pssmdir = '/ifs/home/liudiwei/experiment/step2/pssm/newpssm'
newdir = '/ifs/home/liudiwei/experiment/step2/pssm/computedpssm'
window_size = 5
computedPSSM(window_size,pssmdir, newdir)
'''

平滑的PSSM，只是pssmdir不同，直接调用standardPSSM函数

def smoothedPSSM(window_size,pssmdir,outdir):
    standardPSSM(window_size,pssmdir, outdir)

'''Test example
pssmdir = '/ifs/home/liudiwei/experiment/step2/pssm/computedpssm'
newdir = '/ifs/home/liudiwei/experiment/step2/pssm/smoothedpssm'
window_size = 5
smoothedPSSM(window_size,pssmdir,newdir)
'''

最后得到的是一个滑动的PSSM矩阵，特征的维数随窗口的大小逐渐增减。

---

本栏目python代码分享持续更新中，欢迎关注：Dream_Angel_Z博客 新浪微博： 拾毅者