R包：DiagrammeR流程图

介绍

DiagrammeR 依赖于图形描述语言 Graphviz，可以通过 R 包 igraph 和 visNetwork 访问。DiagrammeR 通过将有效的图规范以 DOT 语言的形式传递给 grViz() 函数来输出图。

加载R包

采用DiagrammeRR包，它提供了以下函数：

使用create_graph()和render_graph()在节点和边的列表上。
创建一个有效的DOT语言图规范，并将此作为字符串传递给函数grViz()。
创建一个有效的图规范，并将此作为字符串或 .mmd 文件引用传递给函数 mermaid()。

{r, 复制代码

knitr::opts_chunk$set(message = FALSE, warning = FALSE)

if (!require("pacman")) {
  install.packages("pacman")
}
pacman::p_load(here,
               tidyverse,
               DiagrammeR,
               DiagrammeRsvg,
               rsvg, 
               xml2,
               plotly)

rm(list = ls())
options(stringsAsFactors = F)
options(future.globals.maxSize = 10000 * 1024^2)

create_graph创建图

简单的图可以通过在节点和边的数据框列表上使用 create_graph() 方法来创建。这种方法对于绘制对象之间的简单关系非常方便。

{r} 复制代码

# 定义图边和节点 
# 1 == "Photosynthesis"
# 2 == "Aerobic respiration"  
# 3 == "Solar energy"  

edges <- create_edge_df(from = c(1, 2, 3), 
                        to = c(2, 1, 1), 
                        label = c("Oxygen \n Glucose", "Carbon dioxide \n Water", NA),
                        color = "black") 

nodes <- create_node_df(n = 3,
                        label = c("Photosynthesis", "Aerobic \n respiration", "Solar \n energy"),
                        type = "lower",
                        style = "filled",
                        fillcolor = c("darkseagreen", "mistyrose", "gold"), 
                        fontcolor = "black",
                        shape = "circle",
                        fixedsize = T)

# 创建图
simple_graph <- create_graph(nodes_df = nodes, edges_df = edges)  

# 渲染图
render_graph(simple_graph)

# 导出图
# export_graph(simple_graph,
#              file_name = here("figures", "dv-using_diagrammer-simple_flowchart.svg"),
#              file_type = "svg")

如果您想创建更复杂的流程图，这种方法存在一些限制。

边必须使用到和从的数值向量进行编码。
没有简单的方法来指定节点位置（将太阳能显示为顶部节点是有意义的）。
没有简单的方法来指定边标签的位置。
如果节点标签太长，图形输出会被修剪。

grVis创建流程图

创建流程图

{r} 复制代码

simple_flowchart <- grViz(
  "digraph {
         
    graph[layout = dot, rankdir = LR] # Set node direction
                     
    node[shape = circle, style = filled] # Set global node attributes 
                     
    a[label = 'Photosynthesis', fillcolor = 'darkseagreen']
    b[label = 'Aerobic \nrespiration', fillcolor = 'mistyrose']
    c[label = 'Solar \nenergy', fillcolor = 'gold']

    edge[color = black, fontsize = 8] # Set global edge attributes  
                     
    a -> b[label = 'Oxygen \nGlucose']
    a -> b[style = 'invis']
    a -> b[style = 'invis']
    b -> a[label = 'Carbon dioxide \n Water']
    b -> a[style = 'invis']
    b -> a[style = 'invis']
    c -> a[label = '                       ']
                     
  }"
)

simple_flowchart

# # Export graph as an svg -------------------------------------------------------
# simple_flowchart %>% 
#   export_svg %>%
#   charToRaw %>%
#   rsvg_svg(here("figures", "dv-using_diagrammer-simple_grvis.svg"))

为节点和边设置全局属性，并在需要时使用局部属性手动覆盖个别节点或边。
按照这里推荐的方法，创建额外的不可见边层以增加特定节点之间的箭头曲率。
创建一个带有空字符串的边标签，以增加特定边的长度。

grVis创建复杂流程图

机器学习流程图

{r} 复制代码

# Create a flow chart to describe my ML workflow -------------------------------  
ml_workflow <- grViz(
  "digraph {
         
    graph[layout = dot]
                     
    node[shape = rectangle, style = filled, fillcolor = 'blanchedalmond'] 
                     
    data_source_1[label = 'Data \n warehouse', shape = cylinder, fillcolor = 'azure']
    data_source_2[label = 'Survey.csv', fillcolor = 'aliceblue']
    data_source_3[label = 'External data \n API', shape = cylinder, fillcolor = 'azure']
                  
    process_1[label = 'Data cleaning \n and joining']
    process_2[label = 'Feature selection']
    process_3[label = 'Split train \n and test dataset']  
    process_4[label = 'Cross-validation']  
    process_5[label = 'Test different \n ML algorithms']
    process_6[label = 'Select optimal \n ML model']
                     
    file_1[label = 'Clean data', fillcolor = 'aliceblue']
    file_2[label = 'Training data', fillcolor = 'aliceblue']
    file_3[label = 'Test data', fillcolor = 'aliceblue']
                     
    product_1[label = 'Optimal \n ML model', shape = 'ellipse', fillcolor = 'lightsalmon']
                     
    edge[color = black, fontsize = 12] 
                     
    data_source_1 -> process_1
    data_source_2 -> process_1
    data_source_3 -> process_1  
                     
    process_1 -> file_1
                     
    file_1 -> process_2
    process_2 -> process_3
    process_3 -> file_2
    process_3 -> file_3
                     
    file_2 -> process_4
    process_4 -> process_5
    process_5 -> process_4
                     
    process_5 -> file_3
    file_3 -> process_6
    process_6 -> product_1
                     
    product_1 -> file_2[label = '      Constantly monitor \n for ML model drift',
                        style = 'dashed', penwidth = 2, weight = 2, 
                        color = 'firebrick', fontcolor = 'firebrick']
                     
  }"
)

ml_workflow


# # Export graph as an svg -------------------------------------------------------
# ml_workflow %>% 
#   export_svg %>%
#   charToRaw %>%
#   rsvg_svg(here("figures", "dv-using_diagrammer-ml_workflow.svg"))

临床试验图

{r} 复制代码

# Create flow chart with flexible parameter inputs -----------------------------  
set.seed(111)

a <- 200 # Total patients
b <- sample(1:60, 1) # Excluded patients
c <- 200 - b # Randomised patients 
d <- ceiling(c/2) # Assigned treatment A
e <- c - d # Assigned treatment B

# Store parameters inside a list
flow_chart_data <- list(a, b, c, d, e) 

# Create grVis() graph
simple_trial <- grViz(
  "digraph {
         
    graph[layout = dot]
                     
    node[shape = rectangle, style = filled, margin = 0.2, fillcolor = 'aliceblue']  
                     
    a[label = '@@1']
    b[label = '@@2', fillcolor = 'mistyrose']
    c[label = '@@3']
    d[label = '@@4']
    e[label = '@@5']

    edge[color = black]   
                
    a -> b[style = 'dashed']
    a -> c[weight = 2] # Weighs this edge more heavily so it is centrally placed 
    c -> d
    c -> e
                     
  }
                     
  [1]: paste0('Patients screened (n = ', flow_chart_data[[1]], ')')
  [2]: paste0('Patients excluded (n = ', flow_chart_data[[2]], ')')
  [3]: paste0('Patients randomised (n = ', flow_chart_data[[3]], ')')
  [4]: paste0('  Patients assigned Treatment A (n = ', flow_chart_data[[4]], ')')
  [5]: paste0('  Patients assigned Treatment B (n = ', flow_chart_data[[5]], ')')

  "
) 

simple_trial

# # Export graph as an svg -------------------------------------------------------
# simple_trial %>% 
#   export_svg %>%
#   charToRaw %>%
#   rsvg_svg(here("figures", "dv-using_diagrammer-flowchart_clinical_single.svg"))

多研究中心临床试验图

{r} 复制代码

# Create flow chart with flexible parameter inputs -----------------------------  
set.seed(111)

a <- 200
b <- sample(1:60, 1)
c <- 200 - b
d <- ceiling(c/2)
e <- c - d

set.seed(222)

f <- 125
g <- sample(1:45, 1)
h <- 125 - b
i <- ceiling(h/2)
j <- h - i

# Store subgroup parameters in separate lists  
hospital_a <- list(a, b, c, d, e) 
hospital_b <- list(f, g, h, i, j)

# Create grVis() graph  
multi_trial <- grViz(
  "digraph {

    graph[layout = 'dot', rankdir = TB]
               
    node[shape = rectangle, style = filled, margin = 0.2, fillcolor = 'azure']
               
    subgraph cluster_a { 
      graph[rankdir = TB, label = 'Hospital A',
            fontsize = 18, shape = rectangle, style = dashed]

      a[label = '@@1']
      b[label = '@@2', fillcolor = 'mistyrose']
      c[label = '@@3']
      d[label = '@@4', fillcolor = 'aliceblue']
      e[label = '@@5', fillcolor = 'aliceblue']
               
    }
               
    subgraph cluster_b { 
      graph[rankdir = TB, label = 'Hospital B',
            fontsize = 18, shape = rectangle, style = dashed]
             
      f[label = '@@6']
      g[label = '@@7', fillcolor = 'mistyrose']
      h[label = '@@8']
      i[label = '@@9', fillcolor = 'aliceblue']
      j[label = '@@10', fillcolor = 'aliceblue']
               
    }
               
    edge[color = black]
               
    a -> b[style = 'dashed']
    a -> c[weight = 2] 
    c -> d
    c -> e
    f -> g[style = 'dashed']
    f -> h[weight = 2] 
    h -> i
    h -> j
               
  }
      
  [1]: paste0('Patients screened (n = ', hospital_a[[1]], ')')
  [2]: paste0('Patients excluded (n = ', hospital_a[[2]], ')')
  [3]: paste0('Patients randomised (n = ', hospital_a[[3]], ')')
  [4]: paste0('  Patients assigned Treatment A (n = ', hospital_a[[4]], ')')
  [5]: paste0('  Patients assigned Treatment B (n = ', hospital_a[[5]], ')')
  [6]: paste0('Patients screened (n = ', hospital_b[[1]], ')')
  [7]: paste0('Patients excluded (n = ', hospital_b[[2]], ')')
  [8]: paste0('Patients randomised (n = ', hospital_b[[3]], ')')
  [9]: paste0('  Patients assigned Treatment A (n = ', hospital_b[[4]], ')')
  [10]: paste0('  Patients assigned Treatment B (n = ', hospital_b[[5]], ')')

  "
)  

multi_trial

# # Export graph as an svg -------------------------------------------------------
# multi_trial %>% 
#   export_svg %>%
#   charToRaw %>%
#   rsvg_svg(here("figures", "dv-using_diagrammer-flowchart_clinical_multiple.svg"))