dump—HiTC

image.png

我们知道.hic 文件是高度压缩的二进制文件，便于存储和分析。那么如果我们想要从.hic提取某一区域的交互信息的话，该如何操作呢？这就涉及到了juicer dump。
https://github.com/aidenlab/juicer/wiki/Data-Extraction
Juicer dump 有以下参数：

Usage:   
juicebox dump    [:x1:x2] [:y1:y2]   [outfile]
         dump       [outfile]
         dump   [outfile]

示例：

juicer_tools dump observed NONE  sam1.chr20.hic   20:32679500:32680500  20 BP 10000  extract_matrix.txt

提取的矩阵主要有三列：（start，end，contacts）
提取矩阵示例：

120000  32680000    1.0
350000  32680000    2.0
370000  32680000    1.0
560000  32680000    1.0
850000  32680000    1.0
980000  32680000    1.0
1190000 32680000    2.0
1270000 32680000    1.0
1300000 32680000    1.0
1800000 32680000    1.0

那么如果我们想要进行可视化的话，可以参照以下代码转换成HiTC格式的矩阵：

#! /usr/bin/env python3
import time
import math
import unittest
import re,sys,os
import numpy as np
import pandas as pd
from scipy import sparse
#step 1  juicer 提取矩阵
#start,end 设定
class trans_hic():

    def __init__(self):

        self.hic=""
        self.start=0
        self.end=100000
        self.juicer_tools=""
        self.outdir=""
        self.bin=10000
        self.genome='hg19'
        self.juicer_dump_mat=""
        self.hitc_matrix=""
        self.prefix="sam1"
    
    def extract_matrix(self): 
        chr=self.chr;start=self.start;end=self.end
        chrom=self.chr.replace('chr','')
        juicer_dump_mat="{}/{}_{}_{}_{}_dump.mat".format(self.outdir,self.prefix,chr,start,end)
        self.juicer_dump_mat=juicer_dump_mat
        region="{0}:{1}:{2} {0}:{1}:{2}".format(chrom,str(start),str(end))
        cmd="/opt/juicer/scripts/juicer_tools dump observed NONE {}  {}  BP {} {}".format(self.hic,region,self.bin,juicer_dump_mat)
        print(cmd)
        os.system(cmd)
    
    def reform_matrix(self): 
        #-----------HiTC matrix---------------------
        chr=self.chr;start=self.start;end=self.end;bin=self.bin;genome=self.genome
        self.hitc_matrix="{}/{}_{}_{}_{}_hitc.mat".format(self.outdir,self.prefix,chr,start,end)
        print('juicer dump matrix....')
        print(self.juicer_dump_mat)
        mat=pd.read_table(self.juicer_dump_mat,names=['frag1','frag2','contacts'])
        #print('matrix head.....')
        #print(mat.head())
        min=math.ceil(int(start)/bin)*bin
        max=int(int(end)/bin)*bin
        N=int(end/bin)-math.ceil(start/bin)+1
        #---------------------- add header --------------------------
        inddf=np.arange(N)
        headers_ref=[genome for x in inddf]
        bin_num_df=pd.Series(inddf).apply(lambda x : str(x))
        headers_ref=pd.Series(headers_ref)
        chromdf=pd.Series([chr for x in list(range(N))])
        startdf=pd.Series(inddf*bin+min)
        enddf=pd.Series((inddf+1)*bin+min)
        headers_suf=chromdf.str.cat(startdf.apply(lambda x :str(x)),sep=':')
        headers_suf=headers_suf.str.cat(enddf.apply(lambda x:str(x)),sep="-")
        headers=bin_num_df.str.cat([headers_ref,headers_suf],sep="|")
        headers=list(headers)

        mat['b1']=mat['frag1'].apply(lambda x: (x-min)/bin)
        mat['b2']=mat['frag2'].apply(lambda x: (x-min)/bin)
        counts=sparse.coo_matrix((mat['contacts'],(mat['b1'],mat['b2'])),shape=(N, N),dtype=float).toarray()
        diag_matrix=np.diag(np.diag(counts))
        counts=counts.T + counts
        counts=counts-diag_matrix-diag_matrix
        df=pd.DataFrame(counts)
        df.columns=headers
        df.index=headers
        #print('DataFrame.....')
        #print(df.head())
        df.to_csv(self.hitc_matrix,sep="\t")
        return df

    def z_score_norm(self):
        print('z-score normlizaion ....................')
        df=self.reform_matrix()
        print('befor zscore.......')
        print(df.head())
        dsc = pd.DataFrame(np.ravel(df)).describe(include=[np.number])
        df = (df - dsc.ix['mean',0])/dsc.ix['std',0]
        print('after zscore....')
        print(df.head())
        return df


class Test_trans(unittest.TestCase):
    def test_trans(self):
        trhic=trans_hic()
        trhic.outdir="/Test"
        trhic.hic="sam1.chr1.hic"
        trhic.chr='chr1'
        trhic.start=62932570
        trhic.end=63564575
        trhic.extract_matrix()
        trhic.reform_matrix()


if __name__ == '__main__':
    unittest.main()

来查看一下结果。

image.png