构造DeBruijn图的算法给出了错误的结果

时间:2015-12-18 11:18:24

标签: python dna-sequence

我正在尝试编写一些代码,用Python中的一组kmers(k字母长字符串,DNA测序读取)构建DeBruijn图,输出为边集合,将同一节点连接到其他节点。

当我在示例输入上运行代码时:

['GAGG','CAGG','GGGG','GGGA','CAGG','AGGG','GGAG']

我明白了:

CAG -> AGG
GAG -> AGG

而不是:

AGG -> GGG
CAG -> AGG,AGG
GAG -> AGG
GGA -> GAG
GGG -> GGA,GGG

任何我做错的提示?
这是代码:

import itertools

inp=['GAGG','CAGG','GGGG','GGGA','CAGG','AGGG','GGAG']

y=[a[1:] for a in inp]
z=[b[:len(b)-1] for b in inp]

y.extend(z)
edjes=list(set(y))

w=[c[1:] for c in edjes]
v=[d[:len(d)-1] for d in edjes]

w.extend(v)

nodes=list(set(w))

graph={}


new=itertools.product(edjes,edjes)

for node in nodes:
    for edj in new:
        edje1,edje2=edj[0],edj[1]
        if edje1[1:]==node and edje2[:len(edje2)-1]==node:
            if edje1 in graph:
                graph[edje1].append(edje2)
            else:
                graph[edje1]=[edje2]

for val in graph.values():
    val.sort()


for k,v in sorted(graph.items()):
    if len(v)<1:
        continue
   else:
        line=k+' -> '+','.join(v)+'\n'
   print (line)

1 个答案:

答案 0 :(得分:3)

我认为你使算法太复杂了:你可以简单地首先对输入执行唯一性过滤器:

inp=['GAGG','CAGG','GGGG','GGGA','CAGG','AGGG','GGAG']

edges=list(set(inp))

然后迭代这个“边缘”列表。对于每个边,前三个字符是来自节点,后三个字符是节点:

for edge in edges:
    frm = edge[:len(edge)-1]
    to = edge[1:]
    #...

现在您只需将其添加到图表中:

for edge in edges:
    frm = edge[:len(edge)-1]
    to = edge[1:]
    if frm in graph:
        graph[frm].append(to)
    else:
        graph[frm]=[to]

最后像你自己一样进行分类和打印:

for val in graph.values():
    val.sort()

for k,v in sorted(graph.items()):
    print(k+' -> '+','.join(v))

这导致:

AGG -> GGG
CAG -> AGG
GAG -> AGG
GGA -> GAG
GGG -> GGA,GGG

正如您所看到的,2行有一个小差异:您的预期输出包含AGG两次,这没什么意义。

所以完整的算法类似于:

inp=['GAGG','CAGG','GGGG','GGGA','CAGG','AGGG','GGAG']

edges=list(set(inp))

graph={}

for edge in edges:
    frm = edge[:len(edge)-1]
    to = edge[1:]
    if frm in graph:
        graph[frm].append(to)
    else:
        graph[frm]=[to]

for val in graph.values():
    val.sort()

for k,v in sorted(graph.items()):
    print(k+' -> '+','.join(v))

您的算法

我认为一个问题是你认为三个字母序列是“edjes”(可能是 edge )。边是四个序列字符。通过执行此转换,信息将丢失。接下来,您构造一组双字符项(nodes,它们根本没有节点)。它们似乎用于将节点“粘合”在一起。但是在那个阶段,你不知道这些碎片是如何粘在一起的。