生信算法2 - DNA测序算法实践之序列统计

发布时间:2023年12月18日

生信序列基本操作算法

建议在Jupyter实践,python版本3.9

1. 读取fastq序列

# fastq序列获取
!wget http://d28rh4a8wq0iu5.cloudfront.net/ads1/data/SRR835775_1.first1000.fastq

def readFastq(filename):
    # 序列列表
    sequences = []
    # 质量值列表
    qualities = []
    with open(filename) as fh:
        # 根据fastq文件格式获取内容
        while True:
            # 跳过行
            fh.readline() 
            # 读取序列,并去除末尾换行符\n
            seq = fh.readline().rstrip() 
            # 跳过行
            fh.readline() 
            # 碱基质量
            qual = fh.readline().rstrip()
            if len(seq) == 0:
                break
            sequences.append(seq)
            qualities.append(qual)
    return sequences, qualities
seqs, quals = readFastq('SRR835775_1.first1000.fastq')
print(seqs)
print(quals)

输出结果

2. 获取fastq数据测序碱基质量直方图

# 质量值转换
def phred33ToQ(qual):
    return ord(qual) - 33

# 获取碱基质量值数据
def createHist(qualities: list):
    hist = [0]* len(qualities)
    for read in qualities:
        for phred in read:
            q = phred33ToQ(phred)
            hist[q] += 1
    return hist
hist_data = createHist(quals)
print(hist_data)

# matplotlib绘图
%matplotlib inline
import matplotlib.pyplot as plt
plt.plot(range(len(hist_data)), hist_data)
plt.show()

碱基质量值直方图

3. 计算fastq序列的平均GC含量

# 获取fastq序列的平均GC含量
def getGCByPosition(reads):
    gc = [0] * 100
    totals = [0] * 100
    for read in reads:
        # 遍历单个reads
        for i in range(len(read)):
            if read[i] == 'C' or read[i] == 'G':
                gc[i] += 1
            # 总reads数加1
            totals[i] += 1
            
    # 获取GC含量平均值
    for i in range(len(gc)):
        if totals[i] > 0:
            gc[i] /= float(totals[i])
    return gc

gc = getGCByPosition(seqs)
# 绘图
plt.plot(range(len(gc)), gc)
plt.show()

GC含量变化图

4. 计算fastq序列的ATCGN碱基数量


import collections
count = collections.Counter()
for seq in seqs:
    count.update(seq)
count
# Counter({'G': 28742, 'C': 28272, 'T': 21836, 'A': 21132, 'N': 18})

5. 读取基因组fasta文件

!wget http://d28rh4a8wq0iu5.cloudfront.net/ads1/data/phix.fa

def readGenome(file_path):
    genome = ''
    with open(file_path, 'r') as f:
        # 遍历fasta文件每行
        for line in f:
            # 忽略>行
            if not line[0] == '>':
                genome += line.rstrip()
    return genome

genome = readGenome('phix.fa')
genome

genome内容

5. 获取短重复序列的序列位置算法

# 获取短重复序列的序列位置
def getRepeatSequencePosition(repeatSeq, seq):
    occurrences = []
    for i in range(len(seq) - len(repeatSeq) + 1):
        match = True
        for j in range(len(repeatSeq)):
            print(seq[i+j], '\t', repeatSeq[j])
            if seq[i+j] != repeatSeq[j]:
                match = False
                break
        if match:
            occurrences.append(f"{i}:{i+j}")
    return occurrences

repeatSeq = 'AG'
seq = 'AGCTTAGATAGCAGG'
getRepeatSequencePosition(repeatSeq, seq)
# ['0:1', '5:6', '9:10', '12:13']

6. 基于基因组序列随机生成reads算法

import random

# 根据给定基因组fasta文件随机生成reads序列
def generateReads(genome, numReads, readLength):
    reads = []
    for _ in range(numReads):
        start = random.randint(0, len(genome)-readLength) - 1
        reads.append(genome[start : start+readLength])
    return reads

# 生成50个长度为100bp的reads序列
reads = generateReads(genome, 50, 100)
reads

reads内容

生信算法文章

生信算法1 - DNA测序算法实践之序列操作

文章来源:https://blog.csdn.net/LittleComputerRobot/article/details/134970714
本文来自互联网用户投稿,该文观点仅代表作者本人,不代表本站立场。本站仅提供信息存储空间服务,不拥有所有权,不承担相关法律责任。