sankey.R

library(tidyverse)
library(patchwork)
library(MetBrewer); library(MoMAColors)

set.seed(81)

setwd("~/Documents/Github/pttp-evidence")
statements <- read_csv("statements.csv")

statements %>% 
  rename(SpeakerType = `Speaker Category`, CitationType = `Category (from Citation)`) -> statements

statements$Speaker <- NA
statements$Speaker[statements$SpeakerType == 'Expert Resource'] <- statements$Individuals[statements$SpeakerType == 'Expert Resource']
statements$Speaker[!(statements$SpeakerType == 'Expert Resource')] <- statements$Organization[!(statements$SpeakerType == 'Expert Resource')]

# Ask this BEFORE separating

prop.table(table(statements$SpeakerType, statements$`Self-citation`), margin = 1)

# separate speakers
statements -> statements.before
statements %>%
  separate_rows(Speaker, sep = ',') -> statements

prop.table(table(statements$SpeakerType, statements$`Self-citation`), margin = 1)

### Clean up self-citation and long names

statements %>%
  mutate(`Self-citation` = recode(`Self-citation`, !!!c('Individual advisor' = "Individual author"))) -> 
  statements

statements %>%
  mutate(Citation = recode(Citation, !!!c("European Commission Joint Research Centre 2024" = "ECJRC 2024")),
         Speaker = recode(Speaker, !!!c("International Disability Alliance" = "Int'l Disability Alliance"))) -> statements

### refactor some things real fast

statements %>%
  select(Speaker, SpeakerType) %>%
  arrange(SpeakerType, Speaker) %>%
  pull(Speaker) %>% unique() -> speaker.levels
statements %>% mutate(Speaker = factor(Speaker, levels = speaker.levels)) -> statements

statements %>%
  select(Citation, CitationType) %>%
  arrange(CitationType, Citation) %>%
  pull(Citation) %>% unique() -> citation.levels
statements %>% mutate(Citation = factor(Citation, levels = citation.levels)) -> statements

statements %>%
  mutate(CitationType = recode(CitationType, !!!c("Journal Article" = "Journal Article, Preprint,\n or Academic Book",
                                                  "Preprint" = "Journal Article, Preprint,\n or Academic Book",
                                                  "Book" = "Journal Article, Preprint,\n or Academic Book",
                                                  "Report or White Paper" = "Technical Report,\n White Paper, or Blog",
                                                  "Editorial or Blog" = "Technical Report,\n White Paper, or Blog",
                                                  "\"IO Guidance, Strategy, or Tool\"" = "International Organization\nGuidance, Strategy, or Tool"))) %>%
  mutate(CitationType = factor(CitationType, levels = c("Journal Article, Preprint,\n or Academic Book",
                                                        "Technical Report,\n White Paper, or Blog",
                                                        "International Organization\nGuidance, Strategy, or Tool"))) %>%
  mutate(SpeakerType = factor(SpeakerType, levels = c('Expert Resource','Stakeholder','Member State'))) -> statements

### Sankey

library(ggalluvial)

statements %>%
  select(SpeakerType, Speaker) %>% distinct() %>%
  pull(SpeakerType) %>% table() -> levs1
statements %>%
  select(CitationType, Citation) %>% distinct() %>%
  pull(CitationType) %>% table() -> levs2
cols <- rev(c("#BD7C96FF", "#D97669FF", "#EEC7A0FF", "#C5EDE1",  "lightblue", "#B5BB8A"))

statements %>%
  group_by(SpeakerType, CitationType) %>%
  summarize(Count = n()) %>% 
  ggplot(aes(axis1 = SpeakerType, axis2 = CitationType, y = Count)) +
  geom_alluvium(aes(fill = `SpeakerType`), width = 0.5) +
  scale_fill_manual(values = c("gray80",'gray50','black')) + 
  ggnewscale::new_scale_fill() + 
  geom_stratum(size = 0.4, width = 0.5) +
  scale_fill_manual(values = cols) +
  geom_text(stat = "stratum",
            aes(label = after_stat(stratum)),
            size = 3.4) +
  theme_void() + 
  theme(legend.position = 'none') -> g1bw; g1bw

statements %>%
  group_by(SpeakerType, CitationType) %>%
  summarize(Count = n()) %>% 
  ggplot(aes(axis1 = SpeakerType, axis2 = CitationType, y = Count)) +
  geom_alluvium(aes(fill = `SpeakerType`), width = 0.5) +
  scale_fill_manual(values = cols[1:3]) + 
  ggnewscale::new_scale_fill() + 
  geom_stratum(aes(fill = after_stat(stratum)), 
               size = 0.4, width = 0.5) +
  scale_fill_manual(values = cols) +
  geom_text(stat = "stratum",
            aes(label = after_stat(stratum)),
            size = 3.4) +
  theme_void() + 
  theme(legend.position = 'none') -> g1color; g1color



statements2 <- statements
statements2 %>%
  select(Speaker, SpeakerType) %>%
  arrange(SpeakerType, Speaker) %>%
  pull(Speaker) %>% unique() -> speaker.levels
statements2 %>% mutate(Speaker = factor(Speaker, levels = speaker.levels)) -> statements2
statements2 %>%
  select(Citation, CitationType) %>%
  arrange(CitationType, Citation) %>%
  pull(Citation) %>% unique() -> citation.levels
statements2 %>% mutate(Citation = factor(Citation, levels = citation.levels)) -> statements2

statements2 %>%
  select(SpeakerType, Speaker) %>% distinct() %>%
  pull(SpeakerType) %>% table() -> levs1
statements2 %>%
  select(CitationType, Citation) %>% distinct() %>%
  pull(CitationType) %>% table() -> levs2

cols <- rev(c("#BD7C96FF", "#D97669FF", "#EEC7A0FF", "#C5EDE1",  "lightblue", ""))

cols <- c(rep("#B5BB8A", levs1[1]),
          rep('lightblue', levs1[2]),
          rep("#C5EDE1", levs1[3]),
          rep("#EEC7A0FF", levs2[1]),
          rep("#D97669FF", levs2[2]),
          rep("#BD7C96FF", levs2[3])
          )

statements2 %>% 
  group_by(SpeakerType, Speaker, CitationType, Citation, `Self-citation`) %>%
  summarize(Count = n()) %>% 
  ggplot(aes(axis1 = Speaker, axis2 = Citation, y = Count)) +
  geom_alluvium(aes(fill = `Self-citation`), alpha = 0.85, width = 0.5) +
  scale_fill_manual(values = c("#B5BB8A",'gray87',"lightblue3")) + 
  ggnewscale::new_scale_fill() + 
  geom_stratum(aes(fill = after_stat(stratum)),
               size = 0.4, width = 0.5) +
  scale_fill_manual(values = cols) +
  geom_text(stat = "stratum",
            aes(label = after_stat(stratum)),
            min.y = 5,
            size = 3) +
  theme_void() + 
  theme(legend.position = 'none') -> g2color; g2color

statements2 %>% 
  group_by(SpeakerType, Speaker, CitationType, Citation, `Self-citation`) %>%
  summarize(Count = n()) %>% 
  ggplot(aes(axis1 = Speaker, axis2 = Citation, y = Count)) +
  geom_alluvium(aes(fill = `Self-citation`), width = 0.5) +
  scale_fill_manual(values = c("black",'gray80',"black")) + 
  ggnewscale::new_scale_fill() + 
  geom_stratum(size = 0.4, width = 0.5) +
  scale_fill_manual(values = cols) +
  geom_text(stat = "stratum",
            aes(label = after_stat(stratum)),
            min.y = 5,
            size = 3) +
  theme_void() + 
  theme(legend.position = 'none') -> g2bw; g2bw

g1color + g2color + plot_annotation(tag_levels = 'A', tag_prefix = "(", tag_suffix = ")") & theme(plot.margin = unit(c(0,0.5,0,0.5),'cm'), plot.tag = element_text(size = 16, hjust = 0, vjust = 0), plot.tag.position = c(-0.05, 0.93))
# g1bw + g2bw + plot_annotation(tag_levels = 'A', tag_prefix = "(", tag_suffix = ")") & theme(plot.margin = unit(c(0,0.5,0,0.5),'cm'), plot.tag = element_text(size = 16, hjust = 0, vjust = 0), plot.tag.position = c(-0.05, 0.93))
ggsave("Figure 1.pdf", width = 13, height = 10)








 ### Network

library(igraph); library(ggraph); library(tidygraph)
library(fs); library(magrittr)

### 81, 53, 46, 43, 37, 26, *****22
### 300, 269, 258, 234, 231, 220, 204
### 199, 198, 174, 151, 150, 105
### 

statements.before %>% 
  group_by(Citation) %>% summarize(Before = n()) -> Before
statements %>% 
  group_by(Citation) %>% summarize(After = n()) -> Fix
Fix %>% left_join(Before) -> Fix


# set.seed(81)

## read in host-virus pairs -----

Edges <- statements %>%
  rename(Mosquito = Speaker, Virus = Citation) %>%
  group_by(Mosquito, Virus) %>% summarize(n = n()) %>%
  group_by(Virus) %>% 
  mutate(nVirus=sum(n)) %>% 
  ungroup() %>% 
  group_by(Mosquito) %>% 
  mutate(nMosquito=sum(n)) %>% 
  ungroup()

# Implement duplication fix

for (i in 1:nrow(Edges)) { 
  Edges$nVirus[i] <- Fix$Before[Fix$Citation==Edges$Virus[i]]
}

## theme ----- 
pal <- MetBrewer::met.brewer('Kandinsky')

## unipartite network ---- 

NodesV <- Edges %>% select(Virus, nVirus) %>% 
  rename(n=nVirus, name=Virus) %>% 
  mutate(NodeCategory="Virus") %>% 
  distinct # keep one row per virus / count - 35 unique 

NodesM <- Edges %>% select(Mosquito, nMosquito) %>% 
  rename(n=nMosquito, name=Mosquito) %>% 
  mutate(NodeCategory="Mosquito") %>% 
  distinct # keep one row per vector / count - 122 unique 

Nodes <- bind_rows(NodesM, NodesV) %>% 
  as.data.frame() %>% 
  mutate(TextSize=sqrt(n)*5) %>% as.data.frame()

Edges %>% 
  select(Mosquito:n) %>% 
  rename(from=Mosquito, to=Virus, weights=n) %>% 
  graph_from_data_frame(., directed = F, Nodes) -> G 

layout = create_layout(G, layout = "dh") %>%
  mutate(name = ifelse(n > 3, name, NA))

cols <- MetBrewer::met.brewer("Cross",5); cols

ggraph(layout) + 
  geom_edge_link(aes(width=weights), alpha=0.5, edge_colour="gray") +
  geom_node_point(aes(size=n, 
                      #shape=NodeCategory, 
                      #alpha=n, 
                      fill=NodeCategory),
                  shape=21, 
                  colour="white", 
                  alpha=0.7) +
  geom_node_text(aes(label=name,
                     colour=NodeCategory,
                     size = TextSize), 
                 repel=TRUE,
                 show.legend = F,
                 alpha=1) + 
  theme_graph() + 
  scale_size(c(0,6))+
  scale_colour_manual(values=c(cols[[4]],cols[[2]])) + 
  scale_fill_manual(values=c(cols[[4]],cols[[2]])) + 
  scale_edge_width(range = c(0.4, 5)) +  # control edge width 
  theme(legend.position = "none")  -> Net; print(Net)

# 
# G %>% as_tbl_graph %>% 
#   ggraph(layout) + 
#   geom_edge_link(aes(width=weights), alpha=0.6, edge_colour="grey") +
#   geom_node_point(aes(size=n, 
#                       #shape=NodeCategory, 
#                       #alpha=n, 
#                       fill=NodeCategory),
#                   shape=21, 
#                   colour= "black", 
#                   alpha=0.8) +
#   geom_node_text(aes(label = name, size=TextSize, 
#                      colour=NodeCategory), 
#                  repel=TRUE,   
#                  show.legend = F) + 
#   theme_graph() + 
#   scale_colour_manual(values=pal[2:1]) + 
#   scale_fill_manual(values=pal[2:1]) + 
#   scale_edge_width(range = c(0.3, 6)) -> NetLegend 

# Network only