Calculate negative effect or competing behaviors in bipartite networks

network-effects.R

library(tidygraph)
library(purrr)
library(dplyr)
library(ggraph)

# Generate sample data: FRAME1
sample_data <- data.frame(genes = c("G1", "G2", "G3", "G4", "G4", "G5", "G6"),
           mirna = c("M1", "M1","M1", "M1", "M2", "M2", "M2"),
           Geneexpression = c(10000, 10000, 5000, 10000, 10000, 5000, 10000),
           mirnaexpression = c(1000, 1000, 1000, 1000, 2000, 2000, 2000), stringsAsFactors = FALSE)
sample_graph <- as_tbl_graph(sample_data)

# Distribute microRNA units to genes proportional to  G1-G6 levels: FRAME2

sample_graph_initial <- sample_graph %N>%
  mutate(type = ifelse(startsWith(name, "G"), "gene", "mirna"), node_id = 1:length(.N()$name)) %E>%
  mutate(Geneexpression_list = as.list(Geneexpression), Geneexpression_pre = Geneexpression, Geneexpression_current = Geneexpression) %>%
  group_by(to) %>%
  mutate(mirnacountpergene = mirnaexpression*Geneexpression_current/sum(Geneexpression_current)) %>%
  ungroup() %>%
  mutate(effect_current = mirnacountpergene, effect_pre = effect_current, effect_list = as.list(effect_current)) 

# G2 is increased : FRAME3
sample_graph_initial %E>%
  mutate(Geneexpression_current = ifelse(from == 2, 20000, Geneexpression_current),Geneexpression_list = pmap(list(Geneexpression_list, Geneexpression_current), c)) -> sample_graph_initial

# keep a copy of graph
result_graph <- sample_graph_initial

# update node values according to edge values
result_graph <- result_graph %N>%
  left_join(bind_rows((as_tibble(result_graph %E>%
                                   select(Geneexpression_current, Geneexpression_pre, mirnaexpression)) %>%
                                   mutate(expression = Geneexpression_pre) %>%
                                   dplyr::select(id = from, expression, expression_pre = Geneexpression_pre,  expression_current = Geneexpression_current) %>%
                                   dplyr::distinct()),
                      (as_tibble(result_graph %E>%
                                   select(mirnaexpression))%>%
                                   mutate(expression = mirnaexpression)%>%
                                   dplyr::select(id= to, expression, expression_pre = mirnaexpression)%>%
                                   dplyr::mutate(expression_current = expression_pre)%>%
                                   dplyr::distinct())), by=c("node_id"="id")) %>%
  mutate(changes_gene = ifelse( expression_current-expression_pre != 0, "Downregulation", type), 
         changes_gene = ifelse(expression_current-expression_pre > 0, "Upregulation", changes_gene)) %>%
  mutate(percent_dif = expression_current/expression)

# For loop that made for recalculation of distribution of microRNAs and expression values of genes
# FRAME4-5-6

for(i in 1:2){

result_graph <- result_graph %E>%
  group_by(to) %>%
  mutate(mirnacountpergene = mirnaexpression*Geneexpression_current/sum(Geneexpression_current)) %>%
  ungroup()%>%
  mutate(effect_pre = effect_current, 
         effect_current = mirnacountpergene)%>%
  group_by(from)%>%
  mutate(Geneexpression_pre = Geneexpression_current, 
         Geneexpression_current = Geneexpression_pre - (sum(effect_current)-sum(effect_pre)))%>%
  ungroup()%>%
  mutate(Geneexpression_list = pmap(list(Geneexpression_list, Geneexpression_current), c), 
         effect_list = pmap(list(effect_list, effect_current), c))
  
result_graph <- result_graph %N>%
  select(-expression_pre)%>%
  select(name, type, node_id, expression, expression_pre = expression_current)%>%
  left_join( bind_rows((as_tibble(result_graph %E>%
                                    select(Geneexpression_current, mirnaexpression))%>%
                                    dplyr::select(id = from, expression_current = Geneexpression_current)%>%
                                    dplyr::distinct()),
                       (as_tibble(result_graph%>%
                                    activate(edges)%>%
                                    select(Geneexpression_current, mirnaexpression))%>%
                                    dplyr::select(id= to,  expression_current = mirnaexpression)%>%
                                    dplyr::distinct())), by=c("node_id"="id"))%>%
  mutate(changes_gene = ifelse( expression_current-expression_pre != 0, "Downregulation", type), 
         changes_gene = ifelse(expression_current-expression_pre > 0, "Upregulation", changes_gene))%>%
  mutate(percent_dif = expression_current/expression)
}

result_graph