We plot the backbone of the flavor network, presented in Nature
The file 'flavor-network.json'
was created from the files provided by the authors, here: https://www.nature.com/articles/srep00196,
and here http://yongyeol.com/pub/.
The json file is a dict with three keys: 'name', 'nodes' and 'edges'.
We also saved in this file the position of each node, assigned by the fdp pygraphviz layout algorithm.
The nodes of the network are the ingredients and node color corresponds to the food category.
Two ingredients are connected if they share a number of flavor compounds.
Hovering the mouse over a node the ingredient name and its prevalence is displayed.
Since for some categories the edges are cluttered we provide the posibility to plot its
corresponding subgraph as a Python igraph with the Kamada-Kaway 3d layout.
import numpy as np
import json
import igraph as ig
import copy
from plotly.offline import download_plotlyjs, init_notebook_mode, iplot
init_notebook_mode(connected=True)
def Plotly_edges(E, node_position):
Xe=[]
Ye=[]
for e in E:
Xe+=[node_position[e[0]][0], node_position[e[1]][0], None]# x coordinates of the nodes defining the edge e
Ye+=[node_position[e[0]][1], node_position[e[1]][1], None]# y
return dict(type='scatter',
x=Xe,
y=Ye,
mode='lines',
line=dict(color='rgb(160,160,160)', width=0.5),
hoverinfo='none',
name='edge'
)
def layout2d(V, node_position):
axis=dict(showbackground=False,#
showline=False,
zeroline=False,
showgrid=False,
showticklabels=False,
title=''
)
layout = dict(title="Backbone of the Flavor Network",
font=dict(family='Balto', color='rgb(5,5,5)'),
width=900,
height=800,
showlegend=True,
xaxis=dict(axis),
yaxis=dict(axis),
margin=dict(t=100, l=20),
hovermode='closest',
legend=dict(x=1., y=0.9),
paper_bgcolor='rgb(235,235,235)',
annotations=[dict(showarrow=False,
text="Data source: <a href='https://www.nature.com/articles/srep00196'> [1]</a>",
xref='paper',
yref='paper',
x=0,
y=-0.05,
xanchor='left',
yanchor='top',
font=dict(family='Balto',
size=14)
)
]
)
anno_text=['flower', 'red_bean', 'onion', 'beef', 'fish', 'cottage_cheese', 'wine',
'olive', 'citrus', 'fennel', 'pepper', 'basil', 'white_bread', 'mushroom']
x_annotext=[node_position[V.index(txt)][0] for txt in anno_text]
y_annotext=[node_position[V.index(txt)][1] for txt in anno_text]
x_anchor=['right', 'right', 'right', 'center', 'center', 'center', 'left',
'center', 'left', 'left', 'right', 'right', 'left', 'left']
y_anchor=['bottom', 'center', 'center', 'bottom', 'top', 'bottom', 'center',
'bottom', 'center', 'center', 'center','bottom', 'center', 'center']
step=4#a value we add/subtract to x_annotext, y_annotext in order to avoid overlaying the annotext and the node dot
for k in range(len(x_anchor)):
if x_anchor[k]=='right': x_annotext[k]-=step
elif x_anchor[k]=='left': x_annotext[k]+=step
else:pass
for k in range(len(y_anchor)):
if y_anchor[k]=='top': y_annotext[k]-=step
elif y_anchor[k]=='bottom': y_annotext[k]+=step
else: pass
layout['annotations']=layout['annotations']+[dict(showarrow=False,
text=anno_text[k],
xref='x',
yref='y',
x=x_annotext[k],
y=y_annotext[k],
font=dict(family='Balto', size=11, color='rgb(5,5,5)'),
xanchor=x_anchor[k],
yanchor=y_anchor[k]
) for k in range(len(anno_text))
]
return layout
#### Process data read from the json file
with open("Data/flavor-network.json") as json_file:
network = json.load(json_file)
#print (network.keys())
V=[net['ingredient'] for net in network['nodes']]#list of nodes
N=len(V)
d_index_ingr=dict(zip(range(N), V))
ingr_preval=[net['prevalence'] for net in network['nodes']]# list of prevalence values for each ingredient(node)
ingr_category={net['ingredient']: net['category'] for net in network['nodes']}# dict ingredient: category
categories=list(set(ingr_category.values()))# list of categories
n_cat=len(categories)
cat_indices=[[V.index(v) for v in V if ingr_category[v]==categories[k]] for k in range(n_cat)]
node_position=[net['position'] for net in network['nodes']]#node position assigned via pygraphviz fdp layout
Edges=[tuple(e['edge']) for e in network['edges']]
E=[(V.index(e[0]), V.index(e[1])) for e in Edges]# edges represented as tuple of linked node indices
colors={'fruit': '#FFF700',
'herb':'rgb(102,194,163)',
'cereal/crop':'rgb(247,231,180)' ,
'flower': '#f500f5',
'spice': '#ffa517',
'alcoholic beverage':'#840123',
'meat': '#d21717',
'animal product':'#d2823b',
'vegetable': '#009b00',
'fish/seafood': '#5aadff',
'plant':'#7b9f20',
'plant derivative': '#acae64',
'dairy':'rgb(250,250,250)',
'nut/seed/pulse': '#8D6645'
}
ingr_color=[colors[ingr_category[v]] for v in V]#list of ingredient(node) colors
edges=Plotly_edges(E, node_position)
nodes=[dict(x=[node_position[idx][0] for idx in cat_indices[k]],
y=[node_position[idx][1] for idx in cat_indices[k]],
mode='markers',
name=categories[k],
marker=dict(symbol='dot',
size=9,
color= colors[categories[k]],
line=dict(color='rgb(140,140,140)', width=0.5)
),
text=[V[idx]+'<br>prevalence='+'{:0.5f}'.format(ingr_preval[idx]) for idx in cat_indices[k]],
hoverinfo='text'
) for k in range(n_cat)
]
data=[edges]+nodes
layout=layout2d(V, node_position)
fig=dict(data=data, layout=layout)
iplot(fig)
In order to plot the 3d graph of a category define the following functions:
def category_data(V, Edges, ingr_category, ingr_preval, d_index_ingr, category):
#prepare data for the subgraph of a category
#categ is a string
index_thiscat=categories.index(category)
subg_Edges=[e for e in Edges if ingr_category[e[0]]==category or ingr_category[e[1]]==category]
subg_E=[(V.index(e[0]), V.index(e[1])) for e in subg_Edges]# edges as tuples of ints (indices of linked nodes
#in the list of global nodes)
subg_nodes=np.array([list(e) for e in subg_E]).reshape(2*len(subg_E))
subg_nodes=np.unique(subg_nodes)# all nodes linked with nodes in this category
nodes_categ=list(set([ingr_category[V[idx]] for idx in subg_nodes]))
nr_cat=len(nodes_categ)
subg_g=ig.Graph()
subg_g.add_vertices(subg_nodes.tolist())
L=len(subg_g.vs)
ig_nodes=dict(zip(subg_g.vs['name'], range(L)))
J= [[n for n in subg_nodes if ingr_category[d_index_ingr[n]]==nodes_categ[k]] for k in range(nr_cat)]
I=[[ig_nodes[n] for n in subg_nodes if ingr_category[d_index_ingr[n]]==nodes_categ[k]] for k in range(nr_cat)]
ig_edges=[(ig_nodes[e[0]], ig_nodes[e[1]]) for e in subg_E]
subg_g.add_edges(ig_edges)
subg_layt=np.array(subg_g.layout('kk3d') )
Xe=[]
Ye=[]
Ze=[]
for e in ig_edges:
Xe+=[subg_layt[e[0]][0], subg_layt[e[1]][0], None]
Ye+=[subg_layt[e[0]][1], subg_layt[e[1]][1], None]
Ze+=[subg_layt[e[0]][2], subg_layt[e[1]][2], None]
subg_edges= [dict(type='scatter3d',
x=Xe,
y=Ye,
z=Ze,
mode='lines',
line=dict(color='rgb(160,160,160)', width=0.75),
hoverinfo='none',
name='edge'
)]
subg_nodes=[dict(type='scatter3d',
x=subg_layt[I[k],0],
y=subg_layt[I[k],1],
z=subg_layt[I[k],2],
mode='markers',
name=nodes_categ[k],
marker=dict(symbol='dot',
size=9,
color= colors[nodes_categ[k]],
line=dict(color='rgb(140,140,140)', width=0.5)
),
text=[V[idx]+'<br>prevalence='+'{:0.5f}'.format(ingr_preval[idx]) for idx in J[k]],
hoverinfo='text'
) for k in range(nr_cat)]
return subg_edges+ subg_nodes
#layout for 3d subnetworks
def layout3d(layout2d, subgraph_title):
noaxis=dict(showbackground=False,
showgrid=False,
showline=False,
showticklabels=False,
ticks='',
title='',
zeroline=False)
layout_cat=copy.deepcopy(layout2d)
layout_cat['annotations'][1:]=[]
layout_cat.update(title=subgraph_title,
width=800,
height=700,
scene=dict(camera = dict(eye=dict(x=1., y=1., z=1.)),
aspectratio=dict(x=1, y=1, z=0.95),
xaxis=dict(noaxis),
yaxis=dict(noaxis),
zaxis=dict(noaxis),
)
)
return layout_cat
def category_graph3d(V, Edges, ingr_category, ingr_preval, d_index_ingr, category='meat'):
data_cat=category_data(V, Edges, ingr_category, ingr_preval, d_index_ingr, category)
title_cat='3D Subgraph of the ingredients in the '+category+' category'
layout_cat=layout3d(layout, title_cat)
fig3d=dict(data=data_cat, layout=layout_cat)
iplot(fig3d)
category_graph3d(V, Edges, ingr_category, ingr_preval, d_index_ingr, category='meat')
category_graph3d(V, Edges, ingr_category, ingr_preval, d_index_ingr, category='fish/seafood')
from IPython.core.display import HTML
def css_styling():
styles = open("./custom.css", "r").read()
return HTML(styles)
css_styling()