Taxonony

NCBI Taxonomy 信息

关联：NR SWissport GO KEGG 已实现相互注释

用法

wget ftp://ftp.ncbi.nih.gov/pub/taxonomy/taxdump.tar.gz
tar -xzf taxdump.tar.gz

1. 下载 taxdump.tar.gz
2. 用法
   • 从names.dmp中找到目标物种的taxid
   • 从nodes.dmp中回溯其parent id直至顶端
   • 利用names.dmp将parent id翻译为物种名
   • 注意，ncbi中可能会有taxa名称相同、但taxid不同的情况，他们就是不同的taxa。
   • 另，使用前，也可以删除一些无意义的字符，比如 ' " [] ()

Script1

将信息展开储存：选用names中储存学名的行，nodes的taxid、parent_id、tax_level列

import pandas as pd
import numpy as np

output_file = 'taxid_anno.txt'
names_dir = 'names.dmp'
nodes_dir = 'nodes.dmp'
abbr_dict = dict(zip(
    ['d','p','c','o','f','g','s'],    
    ['superkingdom','phylum','class','order','family','genus','species']
))


## {'genbank acronym', 'blast name', 'acronym', 'type material', 'common name', 'scientific name', 'synonym', 'genbank common name', 'authority', 'in-part', 'includes', 'equivalent name'}
names_df = pd.read_csv(names_dir,sep = '\t',header = None)[[0,2,6]]
names_df.columns = ['id','Name','Type']
names_df = names_df[names_df['Type'].map(lambda x : 'scientific' in x)]
names_df.set_index('id',inplace = True)
id_name_dict = dict(zip(names_df.index.values,names_df['Name'])) 

nodes_df = pd.read_csv(nodes_dir,sep = '\t',header = None)[[0,2,4]]
nodes_df.columns = ['id','parent_id','level']
nodes_df.set_index('id',inplace = True)
id_parent_dict = dict(zip(nodes_df.index.values,nodes_df['parent_id'])) 
id_level_dict = dict(zip(nodes_df.index.values,nodes_df['level'])) 


def retrieve_ids(taxid,fullid_lst,id_parent_dict):
    fullid_lst.append(taxid)    
    parent = id_parent_dict[taxid]
    if taxid == parent:
        return fullid_lst
    return retrieve_ids(parent,fullid_lst,id_parent_dict)


def retrieve_names(fullid_lst,id_name_dict,id_level_dict):  ## several 'no rank' can be overwriten by the last one: 'root'
    level_name_dict = {}
    for id in fullid_lst:
        if id in id_name_dict.keys():
            level_name_dict[id_level_dict[id]] = id_name_dict[id]
        else:
            level_name_dict[id_level_dict[id]] =  'Unknown'
    return level_name_dict

def abbr_names(level_name_dict,abbr_dict):
    full_str = ''
    for k in abbr_dict:
        if abbr_dict[k] in level_name_dict.keys():
            full_str = full_str + k + '__' + level_name_dict[abbr_dict[k]] + ';'
        else:
            full_str = full_str + k + '__No_Rank;'
    return full_str.strip(';')


with open(output_file,'w') as f:
    for taxid in nodes_df.index.values:
        fullid_templst = retrieve_ids(taxid,[],id_parent_dict)
        level_name_tempdict = retrieve_names(fullid_templst,id_name_dict,id_level_dict)
        full_tempstr = abbr_names(level_name_tempdict,abbr_dict)
        f.write('{}\t{}\n'.format(taxid,full_tempstr))

Script2

本示例中，由于只关注界门纲目科属种(NCBI中的Domain其实是superkingdom)，因此使用dict形式来返回结果，方便提取。若要完全NCBI的taxonomy，改用list或者string就好。

另：uniqName_id_dict_eachlevel存储每个目标level中没有重复的taxa name，若名字重复，只能从names_df里找了，会慢一点。

import pandas as pd
import numpy as np

# grep 'scientific name'  names.dmp > names.dmp.scientific
# cut -f 1,3,5 nodes.dmp > nodes.dmp_simplified
names_dir = 'names.dmp.scientific'
nodes_dir = 'nodes.dmp_simplified'
target_level = ['superkingdom','phylum','class','order','family','genus','species']
target_level_abbrev = ['d','p','c','o','f','g','s']
abbrev_dict = dict(zip(target_level,target_level_abbrev)) 


names_df = pd.read_csv(names_dir,sep = '\t',header = None)[[0,2]]
names_df.columns = ['id','Name']
names_df.set_index('id',inplace = True)
id_names_dict = dict(zip(names_df.index.values,names_df['Name']))                      ##

nodes_df = pd.read_csv(nodes_dir,sep = '\t',header = None)
nodes_df.columns = ['id','parent_id','level']
nodes_df.set_index('id',inplace = True)
id_parentid_dict = dict(zip(nodes_df.index.values,nodes_df['parent_id'].values))       ##
id_level_dict = dict(zip(nodes_df.index.values,nodes_df['level'].values))              ##

uniqName_id_dict_eachlevel = {}
for temp_lvl in target_level:
    temp_all_ids = nodes_df[nodes_df['level'] == temp_lvl].index.values
    temp_names_df = names_df.loc[temp_all_ids].copy()
    temp_names_df['id'] = temp_names_df.index.values
    temp_val_counts = temp_names_df['Name'].value_counts()
    temp_val_counts = temp_val_counts[temp_val_counts == 1]
    temp_val_counts = pd.DataFrame({'Name' : temp_val_counts.index.values})
    temp_uniq_df = pd.merge(temp_names_df,temp_val_counts,on='Name')
    uniqName_id_dict_eachlevel[temp_lvl] = dict(zip(temp_uniq_df['Name'],temp_uniq_df['id']))

def find_tax(taxid,tax_dict,id_names_dict = id_names_dict, id_parentid_dict = id_parentid_dict,id_level_dict=id_level_dict):   
    temp_name = id_names_dict[taxid]
    temp_level = id_level_dict[taxid]
    temp_parent_id = id_parentid_dict[taxid]
    tax_dict[temp_level] = temp_name
    if taxid != temp_parent_id:
        tax_dict = find_tax(temp_parent_id,tax_dict)
    return tax_dict                            # this method will loose 'no rank' layers, since there might be multiple of them!

def tax_dict2str(tax_dict,target_level = target_level):
    tax_str = ''
    for lvl in target_level:
        try:
            val = abbrev_dict[lvl] + '__' + tax_dict[lvl]+ ';'
        except:
            val = abbrev_dict[lvl] + '__' + 'norank' + ';'
        tax_str += val
    tax_str = tax_str.strip(';')
    return tax_str

################### Example 1 #####
e1_name = 'Dickeya phage phiDP10.3'
e1_taxid = uniqName_id_dict_eachlevel['species'][e1_name]
e1_taxdict = find_tax(e1_taxid,{})
e1_taxfull = tax_dict2str(e1_taxdict)