Network analysis
What we can get in Network Analysis from LorMe?
LorMe provides traditional network as well as meta-network across treatment. Additionally, network various network visualization, module analysis were optional.
##Traditional Network For traditional network, at least eight samples are required. Ten to twelve samples are recommended.
### Data preparation
data("Two_group") #contained all Treatment
Treat_obj<- sub_tax_summary(taxobj =Two_group, Group=="Treatment") # select samples from 'Treatment'
Control_obj<- sub_tax_summary(taxobj =Two_group, Group=="Control") # select samples from 'Treatment'
### Analysis
Treat_network_results<- network_analysis(
taxobj = Treat_obj,
taxlevel = "Base",
n = 4,
threshold = 0.6
)## Total degree: 172
## Total edges/links: 86
## Total vertices: 141
## Connectance: 0.008713273
## Average degree: 1.219858
## Diameter: 4
## Average path length: 1.440945
## Global clustering coefficient: 0
## Number of clusters: 57
## Betweenness centralization: 0.0006316987
## Degree centralization: 0.0127153
#### Extract nodes_info
nodes_table <-Treat_network_results$Nodes_info
#### Extract Adjacency
Adjacency_table <-Treat_network_results$Adjacency_column_table
#### Extract Adjacency matirx
Adjacency_matrix <-Treat_network_results$Adjacency_matrix
### Same for Control group
Control_network_results<- network_analysis(
taxobj = Control_obj,
taxlevel = "Base",
n = 4,
threshold = 0.6
)## Total degree: 154
## Total edges/links: 77
## Total vertices: 132
## Connectance: 0.008905852
## Average degree: 1.166667
## Diameter: 3
## Average path length: 1.314286
## Global clustering coefficient: 0
## Number of clusters: 55
## Betweenness centralization: 0.0005620974
## Degree centralization: 0.01399491
Visualization
Treat_visual<- network_visual(
network_obj=Treat_network_results,
mode = "major_module", #Type of visualization, "major_module" or "major_tax".
major_num = 5, #The number of modules to display, by default
taxlevel = NULL, #Taxonomy levels to display,only for "major_tax" mode
select_tax = NULL, #specific taxa to display, only for "major_tax" mode
palette = "Set1", #Palette, by default
vertex.size = 6 #The size of the vertices, by default
)
### same for 'Control'
Control_visual<- network_visual(
network_obj=Control_network_results #Other parameters are set as default
)
Module Analysis
### Module composition
module_results <- Module_composition(
network_obj = Treat_network_results,
No.module = c(2, 6), #Module number to analysis
taxlevel = "Phylum", #summarise level
palette = "Plan10",
rmprefix = "p__"
)
module_results$Module2$Pie_plot_Module2
### Module abundance
module_rel<- Module_abundance(Treat_network_results,No.module = c(2,6))
module_rel$rowframe # not available for visualization here## SampleID Group Rep Module2 Module6
## 1 Sample1 Treatment 1 0.001796945 0.002524280
## 11 Sample2 Treatment 2 0.001782683 0.003862479
## 25 Sample11 Treatment 3 0.002042379 0.002467875
## 26 Sample12 Treatment 4 0.004119071 0.003736889
## 27 Sample13 Treatment 5 0.005723877 0.004475813
## 28 Sample14 Treatment 6 0.001237360 0.001877373
## 29 Sample15 Treatment 7 0.001701331 0.003572796
## 30 Sample16 Treatment 8 0.001488538 0.001786246Meta Network
Meta network constructs co-occurrence network across all samples. Combined with differential analysis, we can evaluate the sensitivity of network module to different treatment.
#### Analysis
network_results <- network_analysis(
taxobj = Two_group,
taxlevel = "Genus",
n = 10,
threshold = 0.8
)## Total degree: 278
## Total edges/links: 139
## Total vertices: 115
## Connectance: 0.02120519
## Average degree: 2.417391
## Diameter: 10
## Average path length: 3.735019
## Global clustering coefficient: 0.2988281
## Number of clusters: 23
## Betweenness centralization: 0.08704612
## Degree centralization: 0.1103738
## Calculating.......#### Visualize
network_diff_obj <- network_withdiff(
network_obj = network_results,
diff_frame = indicator_results
)
## xchar= 0.07946,0.07946,0.07946 ; (yextra, ychar)= 0,0,0, 0.1059,0.1059,0.1059
## points2( 1.285 1.285 1.285 , 1.006 0.9005 0.7946 , pch= 16 16 16 , ...)## No.module sum_tag_number
## 1 1 18
## 4 4 15
## 3 3 7
## 6 6 5
## 8 8 4
## 2 2 2
## 9 9 2
## 13 14 2
## 14 16 2
## 15 17 2
## 16 20 2
## 17 22 2
## 18 23 2
## 5 5 1
## 7 7 1
## 10 10 1
## 11 11 1
## 12 13 1# Check contained different tags for each model
print(network_diff_obj$tag_statistics$detailed_tags)## No.module Control Treatment
## 1 1 12 6
## 2 2 0 2
## 3 3 4 3
## 4 4 12 3
## 5 5 1 0
## 6 6 5 0
## 7 7 1 0
## 8 8 4 0
## 9 9 1 1
## 10 10 0 1
## 11 11 1 0
## 12 13 1 0
## 13 14 2 0
## 14 16 2 0
## 15 17 1 1
## 16 20 2 0
## 17 22 0 2
## 18 23 1 1#### Re-visualize
network_visual_re(
network_visual_obj = network_diff_obj,
module_paint = TRUE,
module_num = c(1, 4)
) # Show module with most Treatment indicators
## xchar= 0.07946,0.07946,0.07946 ; (yextra, ychar)= 0,0,0, 0.1059,0.1059,0.1059
## points2( 1.285 1.285 1.285 , 1.006 0.9005 0.7946 , pch= 16 16 16 , ...)###other types please explore from '?network_withdiff'
#### Module composition analaysis were conducted as above
#### Module abundance analsyis
moduleframe<- Module_abundance(network_obj =network_results,No.module = c(1,4) )## ###Distribution hypothesis####
## Normality Test (Shapiro-Wilk): Failed (P = 0.005569698 )
## Equal Variance Test (Brown-Forsythe): Failed (P = 0.04888781 )
## Treatment_Name N Median Q1 Q3
## 1 Control 8 0.07400502 0.06723764 0.07317343
## 2 Treatment 8 0.11297729 0.09800371 0.13201030
##
##
## ###Dependent Variable: Rel ####
##
## ###Mann-Whitney Rank Sum Test####
##
## ###Statistics####
##
## Mann-Whitney U Statistic= 1
##
## n(small)= 8 , n(big)= 8
## P_estimate = 0.001359376 ,P_exact = 0.001131326
##
## ###Conclusion####
## The difference in the median values between the two groups is greater than would be expected by chance; there is a statistically significant difference (P = 0.001131326 ).
##
## ###Distribution hypothesis####
## Normality Test (Shapiro-Wilk): Passed (P = 0.363 )
##
## Equal Variance Test (Brown-Forsythe): Passed (P = 0.961 )
## ###T-test begin ####
##
## ###Dependent Variable: Rel ####
##
## Treatment_Name N Mean Sd SEM
## 1 Control 8 0.05318120 0.005723841 0.002023683
## 2 Treatment 8 0.02921835 0.007167693 0.002534162
## ###Statistics####
##
## Method: Welch Two Sample t-test (Two-tailed)
##
## data: Rel
## t = 7.389, df = 13.347, p-value = 4.522e-06
## alternative hypothesis: true difference in means between group Control and group Treatment is not equal to 0
## 95 percent confidence interval:
## 0.01697517 0.03095054
## sample estimates:
## mean in group Control mean in group Treatment
## 0.05318120 0.02921835
##
## ###Conclusion####
## The difference in the mean values of the two groups is greater than would be expected by chance;
## There is a statistically significant difference between the input groups (P = 4.521846e-06 )## $Data_description
## Treatment_Name N Median Q1 Q3
## 1 Control 8 0.07400502 0.06723764 0.07317343
## 2 Treatment 8 0.11297729 0.09800371 0.13201030
##
## $Statistics
##
## Wilcoxon rank sum test
##
## data: get(colnames(data)[value_col]) by get(colnames(data)[treatment_col])
## W = 1, p-value = 0.001131
## alternative hypothesis: true location shift is not equal to 0
