In [1]:

    

!python --version

!wget http://geneontology.org/gene-associations/gene_association.sgd.gz -O ./data/gene_association.sgd.gz
!wget http://purl.obolibrary.org/obo/go.obo -O ./data/go.obo


    

    




Python 3.5.3 :: Anaconda 4.4.0 (x86_64)
--2017-08-10 17:17:27--  http://geneontology.org/gene-associations/gene_association.sgd.gz
Resolving geneontology.org... 52.27.86.54
Connecting to geneontology.org|52.27.86.54|:80... connected.
HTTP request sent, awaiting response... 200 OK
Length: 1488961 (1.4M) [application/x-gzip]
Saving to: ‘./data/gene_association.sgd.gz’

./data/gene_associa 100%[===================>]   1.42M  4.88MB/s    in 0.3s    

2017-08-10 17:17:28 (4.88 MB/s) - ‘./data/gene_association.sgd.gz’ saved [1488961/1488961]

--2017-08-10 17:17:28--  http://purl.obolibrary.org/obo/go.obo
Resolving purl.obolibrary.org... 52.3.123.63
Connecting to purl.obolibrary.org|52.3.123.63|:80... connected.
HTTP request sent, awaiting response... 302 Found
Location: http://snapshot.geneontology.org/ontology/go.obo [following]
--2017-08-10 17:17:28--  http://snapshot.geneontology.org/ontology/go.obo
Resolving snapshot.geneontology.org... 54.230.84.41, 54.230.84.147, 54.230.84.215, ...
Connecting to snapshot.geneontology.org|54.230.84.41|:80... connected.
HTTP request sent, awaiting response... 200 OK
Length: 36333288 (35M) [application/rdf+xml]
Saving to: ‘./data/go.obo’

./data/go.obo       100%[===================>]  34.65M  48.5MB/s    in 0.7s    

2017-08-10 17:17:29 (48.5 MB/s) - ‘./data/go.obo’ saved [36333288/36333288]

In [2]:

    

import pandas as pd
from goatools import obo_parser


    

In [3]:

    

# Latest data sources

treeSourceUrl = 'http://chianti.ucsd.edu/~kono/ci/data/collapsed_go.no_IGI.propagated.small_parent_tree'
oboUrl = './data/go.obo'
yeastAnnotationUrl = './data/gene_association.sgd.gz'
kegg2goUrl = 'http://geneontology.org/external2go/kegg2go'
reactome2go = 'http://geneontology.org/external2go/reactome2go'

phenotypeUrl='http://downloads.yeastgenome.org/curation/literature/phenotype_data.tab'


    

In [4]:

    

# Load the tree data
treeColNames = ['parent', 'child', 'type', 'in_tree']
tree = pd.read_csv(treeSourceUrl, delimiter='\t', names=treeColNames)
tree.tail(10)


    

    
Out[4]:






  

    

      

      
parent
      
child
      
type
      
in_tree
    
  
  

    

      
441927
      
GO:0090150
      
YHR083W
      
gene
      
NOT_TREE
    
    

      
441928
      
GO:0005575
      
YHR083W
      
gene
      
NOT_TREE
    
    

      
441929
      
GO:0098796
      
YHR083W
      
gene
      
NOT_TREE
    
    

      
441930
      
GO:1902589
      
YHR083W
      
gene
      
NOT_TREE
    
    

      
441931
      
GO:0044085
      
YHR083W
      
gene
      
NOT_TREE
    
    

      
441932
      
GO:0015031
      
YHR083W
      
gene
      
NOT_TREE
    
    

      
441933
      
GO:1902582
      
YHR083W
      
gene
      
NOT_TREE
    
    

      
441934
      
GO:1902580
      
YHR083W
      
gene
      
NOT_TREE
    
    

      
441935
      
GO:0098799
      
YHR083W
      
gene
      
NOT_TREE
    
    

      
441936
      
GO:0098798
      
YHR083W
      
gene
      
NOT_TREE
    
  

In [5]:

    

cols = pd.read_csv('./annotation_columns.txt', names=['col_names'])
col_names = cols['col_names'].tolist()
print(col_names)


    

    




['DB', 'DB_Object_ID', 'DB_Object_Symbol', 'Qualifier', 'GO_ID', 'DB:Reference', 'Evidence', 'With_or_From', 'Aspect', 'DB_Object_Name', 'DB_Object_Synonym', 'DB_Object_Type', 'taxon', 'Date', 'Assigned_by', 'Annotation_Extension', 'Gene_Product_Form_ID']

In [6]:

    

yeastAnnotation = pd.read_csv(yeastAnnotationUrl, delimiter='\t', comment='!', compression='gzip', names=col_names)
yeastAnnotation.head()


    

    
Out[6]:






  

    

      

      
DB
      
DB_Object_ID
      
DB_Object_Symbol
      
Qualifier
      
GO_ID
      
DB:Reference
      
Evidence
      
With_or_From
      
Aspect
      
DB_Object_Name
      
DB_Object_Synonym
      
DB_Object_Type
      
taxon
      
Date
      
Assigned_by
      
Annotation_Extension
      
Gene_Product_Form_ID
    
  
  

    

      
0
      
SGD
      
S000007287
      
15S_RRNA
      
NaN
      
GO:0005763
      
SGD_REF:S000073641|PMID:6262728
      
IDA
      
NaN
      
C
      
Ribosomal RNA of the small mitochondrial ribos...
      
Q0020|14s rRNA|15S_RRNA_2
      
gene
      
taxon:559292
      
20150612
      
SGD
      
NaN
      
NaN
    
    

      
1
      
SGD
      
S000007287
      
15S_RRNA
      
NaN
      
GO:0032543
      
SGD_REF:S000073641|PMID:6262728
      
IC
      
GO:0005763
      
P
      
Ribosomal RNA of the small mitochondrial ribos...
      
Q0020|14s rRNA|15S_RRNA_2
      
gene
      
taxon:559292
      
20150612
      
SGD
      
NaN
      
NaN
    
    

      
2
      
SGD
      
S000007287
      
15S_RRNA
      
NaN
      
GO:0003735
      
SGD_REF:S000073641|PMID:6262728
      
IC
      
GO:0005763
      
F
      
Ribosomal RNA of the small mitochondrial ribos...
      
Q0020|14s rRNA|15S_RRNA_2
      
gene
      
taxon:559292
      
20150612
      
SGD
      
NaN
      
NaN
    
    

      
3
      
SGD
      
S000007288
      
21S_RRNA
      
NaN
      
GO:0005762
      
SGD_REF:S000073372|PMID:6759872
      
IDA
      
NaN
      
C
      
Mitochondrial 21S rRNA
      
Q0158|21S_rRNA_3|21S_rRNA_4
      
gene
      
taxon:559292
      
20040202
      
SGD
      
NaN
      
NaN
    
    

      
4
      
SGD
      
S000007288
      
21S_RRNA
      
NaN
      
GO:0032543
      
SGD_REF:S000073372|PMID:6759872
      
IMP
      
NaN
      
P
      
Mitochondrial 21S rRNA
      
Q0158|21S_rRNA_3|21S_rRNA_4
      
gene
      
taxon:559292
      
20100715
      
SGD
      
NaN
      
NaN
    
  

In [7]:

    

obo = obo_parser.GODag(oboUrl)


    

    




load obo file ./data/go.obo
./data/go.obo: fmt(1.2) rel(2016-12-06) 47,833 GO Terms

In [8]:

    

obo['GO:1902580'].name


    

    
Out[8]:





'single-organism cellular localization'

In [9]:

    

ph = pd.read_csv(phenotypeUrl, delimiter='\t')
ph.head()


    

    
Out[9]:






  

    

      

      
IMI1
      
not in systematic sequence of S288C
      
IMI1.1
      
S000149345
      
PMID: 26091838|SGD_REF: S000180603
      
classical genetics
      
null
      

      
W303
      
mitochondrial genome maintenance: abnormal
      
.1
      
.2
      
.3
      
.4
    
  
  

    

      
0
      
IMI1
      
not in systematic sequence of S288C
      
IMI1
      
S000149345
      
PMID: 26091838|SGD_REF: S000180603
      
classical genetics
      
null
      

      
W303
      
respiratory metabolism: decreased
      
glycerol (2%)
      
nonfermentable carbon source
      
similar results with ethanol and lactate
      

    
    

      
1
      
IMI1
      
not in systematic sequence of S288C
      
IMI1
      
S000149345
      
PMID: 26091838|SGD_REF: S000180603
      
classical genetics
      
null
      

      
W303
      
mitochondrial morphology: abnormal
      

      

      

      

    
    

      
2
      
IMI1
      
not in systematic sequence of S288C
      
IMI1
      
S000149345
      
PMID: 26091838|SGD_REF: S000180603
      
classical genetics
      
null
      

      
W303
      
viable
      

      

      

      

    
    

      
3
      
MAL62
      
not in systematic sequence of S288C
      
MAL62
      
S000029690
      
PMID: 22669197|SGD_REF: S000149697
      
classical genetics
      
overexpression
      

      
Other
      
fermentative growth: increased
      
maltose
      

      
maltose fementation and leavening ability are ...
      

    
    

      
4
      
MATA1
      
not in systematic sequence of S288C
      
MATA1
      
S000029660
      
PMID: 8065362|SGD_REF: S000039420
      
classical genetics
      
overexpression
      

      
Other (LL20)
      
killer toxin resistance: increased
      

      

      
K. lactis zymocin
      

    
  

In [11]:

    

import json

goTree = {
    'data': {
        'name': 'GO Merged Tree'
    },
    'elements': {
        'nodes': [],
        'edges': []
    }
}

print(json.dumps(goTree, indent=4))


    

    




{
    "elements": {
        "edges": [],
        "nodes": []
    },
    "data": {
        "name": "GO Merged Tree"
    }
}

In [12]:

    

node_set= set()

row = next(tree.iterrows())[1]
print(row)
print(row['parent'])


    

    




parent     GO:0046434
child      GO:0009395
type             is_a
in_tree          TREE
Name: 0, dtype: object
GO:0046434

In [13]:

    

def get_node(id, name):
    node = {
        'data': {
            'id': id
        }
    }
    
    if id in obo.keys():
        go = obo[id]
        node['data']['name'] = go.name
        node['data']['namespace'] = go.namespace
        
    return node

def get_edge(source, target, itr, is_tree):
    edge = {
        'data': {
            'source': source,
            'target': target,
            'interaction': itr,
            'is_tree': is_tree
        }
    }
    
    return edge


    

In [14]:

    

edges = []
for data in tree.iterrows():
    row=data[1]
    node_set.add(row['parent'])
    node_set.add(row['child'])
    edges.append(get_edge(row['parent'], row['child'], row['type'], row['in_tree']))


    

In [ ]:

    

nodes = []

for id in node_set:
    node = get_node(id)
    nodes.append(node)


    

In [22]:

    

print(len(nodes))
print(nodes[0])
print(len(edges))
edges[0]


    

    




13037
{'data': {'namespace': 'biological_process', 'id': 'GO:0043648', 'name': 'dicarboxylic acid metabolic process'}}
441937






    
Out[22]:





{'data': {'interaction': 'is_a',
  'is_tree': 'TREE',
  'source': 'GO:0046434',
  'target': 'GO:0009395'}}

In [23]:

    

goTree['elements']['nodes'] = nodes
goTree['elements']['edges'] = edges

with open('./data/tree.cyjs', 'w') as outfile:
    json.dump(goTree, outfile)


    

In [ ]:

    

yeastAnnotation['DB_Object_Synonym'] = yeastAnnotation['DB_Object_Synonym'].fillna('')


    

In [46]:

    

result = yeastAnnotation[yeastAnnotation['DB_Object_Synonym'].str.contains('YHR083W')]
result


    

    
Out[46]:






  

    

      

      
DB
      
DB_Object_ID
      
DB_Object_Symbol
      
Qualifier
      
GO_ID
      
DB:Reference
      
Evidence
      
With_or_From
      
Aspect
      
DB_Object_Name
      
DB_Object_Synonym
      
DB_Object_Type
      
taxon
      
Date
      
Assigned_by
      
Annotation_Extension
      
Gene_Product_Form_ID
    
  
  

    

      
76224
      
SGD
      
S000001125
      
SAM35
      
NaN
      
GO:0005739
      
SGD_REF:S000117178|PMID:16823961
      
IDA
      
NaN
      
C
      
Component of the sorting and assembly machiner...
      
YHR083W|FMP20|SAM complex subunit SAM35|TOB38|...
      
gene
      
taxon:559292
      
20061212
      
SGD
      
NaN
      
NaN
    
    

      
76225
      
SGD
      
S000001125
      
SAM35
      
NaN
      
GO:0003674
      
SGD_REF:S000069584
      
ND
      
NaN
      
F
      
Component of the sorting and assembly machiner...
      
YHR083W|FMP20|SAM complex subunit SAM35|TOB38|...
      
gene
      
taxon:559292
      
20090409
      
SGD
      
NaN
      
NaN
    
    

      
76226
      
SGD
      
S000001125
      
SAM35
      
NaN
      
GO:0005741
      
SGD_REF:S000076328|PMID:15067005
      
IDA
      
NaN
      
C
      
Component of the sorting and assembly machiner...
      
YHR083W|FMP20|SAM complex subunit SAM35|TOB38|...
      
gene
      
taxon:559292
      
20040909
      
SGD
      
NaN
      
NaN
    
    

      
76227
      
SGD
      
S000001125
      
SAM35
      
NaN
      
GO:0001401
      
SGD_REF:S000076328|PMID:15067005
      
IDA
      
NaN
      
C
      
Component of the sorting and assembly machiner...
      
YHR083W|FMP20|SAM complex subunit SAM35|TOB38|...
      
gene
      
taxon:559292
      
20040909
      
SGD
      
NaN
      
NaN
    
    

      
76228
      
SGD
      
S000001125
      
SAM35
      
NaN
      
GO:0045040
      
SGD_REF:S000076328|PMID:15067005
      
IMP
      
NaN
      
P
      
Component of the sorting and assembly machiner...
      
YHR083W|FMP20|SAM complex subunit SAM35|TOB38|...
      
gene
      
taxon:559292
      
20040909
      
SGD
      
NaN
      
NaN
    
    

      
76229
      
SGD
      
S000001125
      
SAM35
      
NaN
      
GO:0005739
      
SGD_REF:S000175707|PMID:24769239
      
IDA
      
NaN
      
C
      
Component of the sorting and assembly machiner...
      
YHR083W|FMP20|SAM complex subunit SAM35|TOB38|...
      
gene
      
taxon:559292
      
20150803
      
SGD
      
NaN
      
NaN
    
    

      
76230
      
SGD
      
S000001125
      
SAM35
      
NaN
      
GO:0001401
      
SGD_REF:S000076561|PMID:15205677
      
IDA
      
NaN
      
C
      
Component of the sorting and assembly machiner...
      
YHR083W|FMP20|SAM complex subunit SAM35|TOB38|...
      
gene
      
taxon:559292
      
20040707
      
SGD
      
NaN
      
NaN
    
    

      
76231
      
SGD
      
S000001125
      
SAM35
      
NaN
      
GO:0001401
      
SGD_REF:S000076561|PMID:15205677
      
IMP
      
NaN
      
C
      
Component of the sorting and assembly machiner...
      
YHR083W|FMP20|SAM complex subunit SAM35|TOB38|...
      
gene
      
taxon:559292
      
20040707
      
SGD
      
NaN
      
NaN
    
    

      
76232
      
SGD
      
S000001125
      
SAM35
      
NaN
      
GO:0005741
      
SGD_REF:S000076561|PMID:15205677
      
IDA
      
NaN
      
C
      
Component of the sorting and assembly machiner...
      
YHR083W|FMP20|SAM complex subunit SAM35|TOB38|...
      
gene
      
taxon:559292
      
20040707
      
SGD
      
NaN
      
NaN
    
    

      
76233
      
SGD
      
S000001125
      
SAM35
      
NaN
      
GO:0006810
      
SGD_REF:S000148669
      
IEA
      
UniProtKB-KW:KW-0813
      
P
      
Component of the sorting and assembly machiner...
      
YHR083W|FMP20|SAM complex subunit SAM35|TOB38|...
      
gene
      
taxon:559292
      
20161112
      
UniProt
      
NaN
      
NaN
    
    

      
76234
      
SGD
      
S000001125
      
SAM35
      
NaN
      
GO:0016020
      
SGD_REF:S000148669
      
IEA
      
UniProtKB-KW:KW-0472
      
C
      
Component of the sorting and assembly machiner...
      
YHR083W|FMP20|SAM complex subunit SAM35|TOB38|...
      
gene
      
taxon:559292
      
20161112
      
UniProt
      
NaN
      
NaN
    
    

      
76235
      
SGD
      
S000001125
      
SAM35
      
NaN
      
GO:0015031
      
SGD_REF:S000148669
      
IEA
      
UniProtKB-KW:KW-0653
      
P
      
Component of the sorting and assembly machiner...
      
YHR083W|FMP20|SAM complex subunit SAM35|TOB38|...
      
gene
      
taxon:559292
      
20161112
      
UniProt
      
NaN
      
NaN
    
    

      
76236
      
SGD
      
S000001125
      
SAM35
      
NaN
      
GO:0005739
      
SGD_REF:S000148669
      
IEA
      
UniProtKB-KW:KW-0496
      
C
      
Component of the sorting and assembly machiner...
      
YHR083W|FMP20|SAM complex subunit SAM35|TOB38|...
      
gene
      
taxon:559292
      
20161112
      
UniProt
      
NaN
      
NaN
    
    

      
76237
      
SGD
      
S000001125
      
SAM35
      
NaN
      
GO:0070096
      
SGD_REF:S000076328|PMID:15067005
      
IMP
      
NaN
      
P
      
Component of the sorting and assembly machiner...
      
YHR083W|FMP20|SAM complex subunit SAM35|TOB38|...
      
gene
      
taxon:559292
      
20090409
      
SGD
      
NaN
      
NaN
    
    

      
76238
      
SGD
      
S000001125
      
SAM35
      
NaN
      
GO:0070096
      
SGD_REF:S000126649|PMID:18039934
      
IMP
      
NaN
      
P
      
Component of the sorting and assembly machiner...
      
YHR083W|FMP20|SAM complex subunit SAM35|TOB38|...
      
gene
      
taxon:559292
      
20090409
      
SGD
      
NaN
      
NaN
    
    

      
76239
      
SGD
      
S000001125
      
SAM35
      
NaN
      
GO:0070096
      
SGD_REF:S000076561|PMID:15205677
      
IMP
      
NaN
      
P
      
Component of the sorting and assembly machiner...
      
YHR083W|FMP20|SAM complex subunit SAM35|TOB38|...
      
gene
      
taxon:559292
      
20090409
      
SGD
      
NaN
      
NaN
    
    

      
76240
      
SGD
      
S000001125
      
SAM35
      
NaN
      
GO:0045040
      
SGD_REF:S000077462|PMID:15326197
      
IMP
      
NaN
      
P
      
Component of the sorting and assembly machiner...
      
YHR083W|FMP20|SAM complex subunit SAM35|TOB38|...
      
gene
      
taxon:559292
      
20090409
      
SGD
      
NaN
      
NaN
    
    

      
76241
      
SGD
      
S000001125
      
SAM35
      
NaN
      
GO:0005741
      
SGD_REF:S000148671
      
IEA
      
UniProtKB-SubCell:SL-0172
      
C
      
Component of the sorting and assembly machiner...
      
YHR083W|FMP20|SAM complex subunit SAM35|TOB38|...
      
gene
      
taxon:559292
      
20161112
      
UniProt
      
NaN
      
NaN
    
    

      
76242
      
SGD
      
S000001125
      
SAM35
      
NaN
      
GO:0045040
      
SGD_REF:S000126649|PMID:18039934
      
IMP
      
NaN
      
P
      
Component of the sorting and assembly machiner...
      
YHR083W|FMP20|SAM complex subunit SAM35|TOB38|...
      
gene
      
taxon:559292
      
20090409
      
SGD
      
NaN
      
NaN
    
    

      
76243
      
SGD
      
S000001125
      
SAM35
      
NaN
      
GO:0005741
      
SGD_REF:S000148669
      
IEA
      
UniProtKB-KW:KW-1000
      
C
      
Component of the sorting and assembly machiner...
      
YHR083W|FMP20|SAM complex subunit SAM35|TOB38|...
      
gene
      
taxon:559292
      
20161112
      
UniProt
      
NaN
      
NaN
    
    

      
76244
      
SGD
      
S000001125
      
SAM35
      
NaN
      
GO:0005741
      
SGD_REF:S000114057|PMID:16407407
      
IDA
      
NaN
      
C
      
Component of the sorting and assembly machiner...
      
YHR083W|FMP20|SAM complex subunit SAM35|TOB38|...
      
gene
      
taxon:559292
      
20060317
      
SGD
      
NaN
      
NaN