Microbial Network Analysis:
Comparing Node Degree Based on Source Feature
Samuel L. Peoples
Background
Jeff Gordon’s A Core Gut Microbiome in Obese and Lean Twins, drew a handful of
conclusions based on a source feature, obesitycat, comparing the microbial
communities of Obese, and Lean individuals. These include:
● Wide array of shared genes; there exists a core microbiome at the gene level.
● Obesity is associated with phylum-level changes in the microbiota.
● Deviations from this core microbiome are associated with physiological states.
The Network
Using an input .biom and mapping file, make_otu_network.py was used to generate the
Cytoscape network files, which were configured as such:
● Edge-Weighted Spring-Embedded Layout
● Blue: Lean, Red: Obese, Pink: OTUs
● Purple lines: Edges
● Node Size: Weighted Degree
Hypothesis
In a network of samples labeled Obese or Lean, connected to nodes representing
OTUs, when measuring the degree of OTUs associated with Obese nodes only,
compared to those associated with Lean-only, will have a higher mean degree, which
represents the higher diversity of OTUs outside the core microbiome.
Generally:
● In a network of samples and OTUs, a difference of the mean degree of OTUs
associated with a boolean category is representative of a deviation from the core
microbiome.
Simple Test Case:
The Script : Usage
network_analysis.py
-node {PATH to NODE FILE}
-edge {PATH to EDGE FILE}
[-o {PATH to OUTPUT DIRECTORY}]
-f {FEATURE COLUMN for comparison}
-c {CATEEGORY of FEATURE} {CATEGORY of FEATURE}
[-n {N_ITERATIONS for Monte Carlo Simulation}]
Under the Hood
Node File
Under the Hood
Edge File
Under the Hood
Join, and separate
● cat_zero_table
● cat_one_table
● df_union
Under the Hood
Further Separation
● otu_zero_only
● otu_one_only
● otu_both
● df_union
Under the Hood
Reduction
otu_zero_only otu_zero_only
otu_one_only otu_one_only
otu_both otu_both
Under the Hood
Sample Node Degree Stats
cat_zero_table
Degree Statistics Dataframe
cat_one_table
{
Sample, Calculate Summary, Append
1000 iterations
}
Under the Hood
OTU Node Degree Stats
otu_zero_only
Degree Statistics Dataframe
otu_one_only
{
Sample, Calculate Summary, Append
1000 iterations
}
otu_both
Under the Hood
Degree Statistics Dataframe Output Statistics
● Minimum
{ Sample, Calculate Summary, Append
1000 iterations
} ●
●
●
●
●
Q1
Mean
Median
Q3
Max
● Std Dev
column.mean()
Test Case Network
Under the Hood
Output statistics
Minimum Q1 Mean Median Q3 Max Std Dev
cat_zero 1.0 1.0 1.396 1.095 1.967 2.0 0.489
cat_one 1.0 1.155 1.669 1.981 2.0 2.0 0.47
otu_zero 1.0 1.0 1.0 1.0 1.0 1.0 0.0
otu_one 1.0 1.0 1.331 1.015 1.841 2.0 0.47
otu_both 2.0 2.0 2.0 2.0 2.0 2.0 0.0
Obese vs. Lean Output
Output statistics
Minimum Q1 Mean Median Q3 Max Std Dev
Lean 206.103 248.114 271.906 271.788 296.252 344.526 35.574
Obese 163.965 246.257 275.77 280.659 308.187 363.067 46.73
otu_Lean 1.0 1.0 1.435 1.01 1.809 4.185 0.759
otu_Obese 1.0 1.003 2.732 1.644 3.016 16.614 3.154
otu_Both 2.067 6.385 24.117 14.049 31.4 119.475 26.959
Obese nodes have a wider distribution of OTU connections, and thus a more
diverse gut microbiome.
OTUs associated with Lean-only nodes have a low mean degree when
compared to those of Obese-only nodes.
Conclusion
● The degree of OTUs associated with Obese-only nodes have a higher degree
when compared to those of Lean-only nodes. This supports the notion of
deviations from the core microbiome being associated with physiological states;
in this case, Obesity.
