BIOPAX Dataset basic metrics computation¶
We demonstarte here how to use Patterns to compute simple metrics ro compare BIOPAX datasets
In [1]:
import logging
logging.getLogger().setLevel(logging.INFO)
In [2]:
from biopax_explorer.pattern.rack import Rack
from biopax_explorer.pattern.pattern import PatternExecutor, Pattern
from biopax_explorer.query import EntityNode
from biopax_explorer.biopax.utils import gen_utils as gu
from biopax_explorer.biopax import *
import matplotlib.pyplot as plt
plt.set_loglevel("info")
import pathlib
import numpy as np
In [3]:
#reactome biopax file level 3 : https://reactome.org/download-data
#imported in fuseki triple store that serves them as sparql endpoints
dataset1 = "reactome_homo"
dataset2 = "reactome_mus"
dataset1_label = "Homo sapiens"
dataset2_label = "Mus musculus"
db = "http://db:3030" # with local triple store deployed with docker-compose
r = Rack()
pe1 = PatternExecutor(db,dataset1) # create a Pattern executor for a dataset
pe2 = PatternExecutor(db,dataset2)
In [4]:
def entitiesByClass(obj):
p=Pattern()
en = EntityNode("ENTITY", obj)
p.define(en)
return p
def definePatterns():
objlist=[Protein(),Complex(),SmallMolecule(),
Control(),BiochemicalReaction()]
return definePatternsFromObj(objlist)
def definePatteyBiosource():
objlist=[BioSource]
return definePatternsFromObj(objlist)
def definePatternsFromObj(objlist):
pdict={}
for obj in objlist:
pdict[obj.__class__.__name__]=entitiesByClass(obj)
return pdict
def countSummarize(result):
uniqbyuri={}
i=0
for entity_row in result:
for entity in entity_row:
i=i+1
#print("%s uri:%s" %(i,entity.pk))
uniqbyuri[entity.pk]=1
return len(uniqbyuri.keys())
In [5]:
def executeQueriesByDataset(pe,patternMap):
counts_ds={}
for clsn, p in patternMap.items():
print(""" execute pattern on %s""" %(clsn))
result = pe.executePattern(p)
count=countSummarize(result)
counts_ds[clsn]=count
return counts_ds
In [6]:
print(""" ds1""",dataset1)
patternMap=definePatterns()
query_counts_ds1=executeQueriesByDataset(pe1,patternMap)
print(query_counts_ds1)
print("""#------------------""")
ds1 reactome_homo
execute pattern on Protein
execute pattern on Complex
execute pattern on SmallMolecule
execute pattern on Control
execute pattern on BiochemicalReaction
{'Protein': 31649, 'Complex': 15700, 'SmallMolecule': 5196, 'Control': 8838, 'BiochemicalReaction': 14709}
#------------------
In [7]:
print(""" ds2""",dataset2)
patternMap=definePatterns()
query_counts_ds2=executeQueriesByDataset(pe2,patternMap)
print(query_counts_ds2)
print("""#------------------""")
ds2 reactome_mus
execute pattern on Protein
execute pattern on Complex
execute pattern on SmallMolecule
execute pattern on Control
execute pattern on BiochemicalReaction
{'Protein': 23605, 'Complex': 11216, 'SmallMolecule': 4460, 'Control': 5906, 'BiochemicalReaction': 9615}
#------------------
In [8]:
# Sample datasets with counts for each query
#query_counts_ds1 = {'Protein': 31649, 'Complex': 15700, 'SmallMolecule': 5196, 'Control': 8838, 'BiochemicalReaction': 14709, 'Gene': 0}
#query_counts_ds2 = {'Protein': 31649, 'Complex': 15700, 'SmallMolecule': 5196, 'Control': 8838, 'BiochemicalReaction': 14709, 'Gene': 0}
In [9]:
def define_figure(query_counts_ds1,query_counts_ds2,
dataset1_label,dataset2_label,xlabel,ylabel, title):
# Extracting queries and counts for each dataset
queries = list(query_counts_ds1.keys())
counts_ds1 = list(query_counts_ds1.values())
counts_ds2 = list(query_counts_ds2.values())
# Setting the width of the bars
bar_width = 0.35
# Creating index for the x-axis
index = np.arange(len(queries))
# Plotting
plt.figure(figsize=(10, 6))
# Plotting bars for dataset DS1
plt.bar(index, counts_ds1, bar_width, label=dataset1_label, color='skyblue')
# Plotting bars for dataset DS2
plt.bar(index + bar_width, counts_ds2, bar_width, label=dataset2_label, color='orange')
# Adding labels and titles
plt.xlabel(xlabel, fontsize=18)
plt.ylabel(ylabel, fontsize=18)
plt.title(title, fontsize=18)
plt.xticks(index + bar_width / 2, queries, rotation=45)
# Adding legend
plt.legend( fontsize=18)
plt.xticks(fontsize=16) # Adjust the font size of x-axis labels
plt.yticks(fontsize=16) # Adjust the font size of y-axis labels
plt.tight_layout()
return plt
In [10]:
exfile="data/output/metrics_simple.pdf"
xlabel="BioPAX classes"
ylabel="Count of Matches"
title='Number of BioPAX entities in Reactome Datasets'
plt=define_figure(query_counts_ds1,query_counts_ds2,
dataset1_label,dataset2_label,xlabel,ylabel,title)
# Save the plot as a PDF file
plt.savefig(exfile)
In [ ]:
In [ ]:
In [11]:
dataset_list=[dataset1,dataset2]
In [12]:
parentClasses=["PhysicalEntity","EntityReference","Interaction"]
sma={}
for cl, children in gu.classes_children().items():
if cl in parentClasses:
sma[cl.lower()]=1
for child in children:
sma[child.lower()]=1
selectedClasses=list(sma.keys())
In [13]:
def count_uniq(matches,pmetalabels):
global selectedClasses
uniq={}
for entity_row in matches:
for entity in entity_row:
if entity.cls.lower() in selectedClasses:
if entity.meta_label is not None and entity.meta_label in pmetalabels:
if entity.pk in uniq.keys():
uniq[entity.pk]=uniq[entity.pk]+1
else:
uniq[entity.pk]=1
return len(list(uniq.keys()))
In [14]:
def prp(data,pe,p,label,pmetalabels):
res = pe.executePattern(p)
#data={label:{"matches":count_uniq(res,pmetalabels)}}
data[label]=count_uniq(res,pmetalabels)
print(data)
querylist=pe.queries(p)
# for q in querylist:
# print("#---generated sparql query---\n\n")
# print(q)
return data
In [15]:
ds_data={}
for dataset in dataset_list:
data={}
pe = PatternExecutor(db,dataset) # create a Pattern executor for a dataset
data= prp(data,pe,r.controlsStateChange(),"controls-state-change-of",["C1"])
#data= prp(data,pe,r.stateChange(),"state-change",["C1","c"])
data=prp(data,pe,r.controlsPhosphorylation(),"controls-phosphorylation",["C1"])
data=prp(data,pe,r.reactsWith(),"reacts-with",["SMR1"])
data=prp(data,pe,r.usedToProduce(),"used-to-produce",["BR"])
data=prp(data,pe,r.controlsExpressionWithTemplateReac(),"controls-expression-of",["Control"])
data=prp(data,pe,r.inComplexWith(),"in-complex-with",["COMPLEX"])
data=prp(data,pe,r.controlsMetabolicCatalysis(),"control-metabolic-catalysis",["Control"])
ds_data[dataset]=data
{'controls-state-change-of': 1616}
{'controls-state-change-of': 1616, 'controls-phosphorylation': 707}
{'controls-state-change-of': 1616, 'controls-phosphorylation': 707, 'reacts-with': 104}
{'controls-state-change-of': 1616, 'controls-phosphorylation': 707, 'reacts-with': 104, 'used-to-produce': 4281}
{'controls-state-change-of': 1616, 'controls-phosphorylation': 707, 'reacts-with': 104, 'used-to-produce': 4281, 'controls-expression-of': 5}
{'controls-state-change-of': 1616, 'controls-phosphorylation': 707, 'reacts-with': 104, 'used-to-produce': 4281, 'controls-expression-of': 5, 'in-complex-with': 3026}
{'controls-state-change-of': 1616, 'controls-phosphorylation': 707, 'reacts-with': 104, 'used-to-produce': 4281, 'controls-expression-of': 5, 'in-complex-with': 3026, 'control-metabolic-catalysis': 4462}
{'controls-state-change-of': 1316}
{'controls-state-change-of': 1316, 'controls-phosphorylation': 540}
{'controls-state-change-of': 1316, 'controls-phosphorylation': 540, 'reacts-with': 96}
{'controls-state-change-of': 1316, 'controls-phosphorylation': 540, 'reacts-with': 96, 'used-to-produce': 3341}
{'controls-state-change-of': 1316, 'controls-phosphorylation': 540, 'reacts-with': 96, 'used-to-produce': 3341, 'controls-expression-of': 32}
{'controls-state-change-of': 1316, 'controls-phosphorylation': 540, 'reacts-with': 96, 'used-to-produce': 3341, 'controls-expression-of': 32, 'in-complex-with': 2542}
{'controls-state-change-of': 1316, 'controls-phosphorylation': 540, 'reacts-with': 96, 'used-to-produce': 3341, 'controls-expression-of': 32, 'in-complex-with': 2542, 'control-metabolic-catalysis': 3424}
In [16]:
print(ds_data)
{'reactome_homo': {'controls-state-change-of': 1616, 'controls-phosphorylation': 707, 'reacts-with': 104, 'used-to-produce': 4281, 'controls-expression-of': 5, 'in-complex-with': 3026, 'control-metabolic-catalysis': 4462}, 'reactome_mus': {'controls-state-change-of': 1316, 'controls-phosphorylation': 540, 'reacts-with': 96, 'used-to-produce': 3341, 'controls-expression-of': 32, 'in-complex-with': 2542, 'control-metabolic-catalysis': 3424}}
In [17]:
subset={}
selectedIndex=[0,1,3,5,6,7,8,9]
for ds, mapd in ds_data.items():
subset[ds]={}
i=-1
for n,ct in mapd.items():
i=i+1
if i in selectedIndex:
subset[ds][n]=ct
In [18]:
exfile="data/output/metrics_adv.pdf"
xlabel="Pattern occurences count"
ylabel="Count of Matches"
title='Number of Pattern matches in Reactome Datasets'
plt=define_figure(subset[dataset1],subset[dataset2],
dataset1_label,dataset2_label,xlabel,ylabel,title)
# Save the plot as a PDF file
plt.savefig(exfile)
In [ ]:
In [ ]: