请求绘制暴露-中介-结局通路网络关系可视化图

需求说明 #

我有一组展示暴露→中介→结局关系的数据，包含1个暴露因素（体力活动）、多个炎症相关中介因素及1个结局（乳腺癌）。需要绘制路径可视化图，满足以下要求：

• 箭头按direction变量（正相关/负相关/无关联）着色
• 箭头粗细随exposure_int_es的绝对值调整
• 暴露节点标注为“Exposure”，结局节点标注为“Outcome”，中介节点保留原名称
• 箭头上标注效应量及95%CI文本

可复现数据集与关联方向判定代码 #

# 可复现数据集
data <- read.table(
  text = "Author Exposure Outcome Pathway exp_in_es_measure exposure_int_es exposure_int_lower exposure_it_higher exp_int_grade exp_int_n_studies
  Swain_CT_2023 Physical_activity CRP Inflammation SMD -0.27 -0.62 0.08 Low 12
  Swain_CT_2023 Physical_activity TNF_alpha Inflammation SMD -0.63 -1.04 -0.22 Moderate 8
  Swain_CT_2023 Physical_activity IL-6 Inflammation SMD -0.55 -0.97 -0.13 Moderate 11
  Swain_CT_2023 Physical_activity Adiponectin Inflammation SMD 0.01 -0.14 0.17 High 5
  Swain_CT_2023 Physical_activity Leptin Inflammation SMD -0.5 -1.1 0.09 Low 4
  Lou_MWC_2023 CRP Breast_Cancer Inflammation RR 1.13 1.01 1.26 Moderate 16
  Lou_MWC_2023 TNF_alpha Breast_Cancer Inflammation RR 0.84 0.57 1.1 Moderate 4
  Lou_MWC_2023 IL-6 Breast_Cancer Inflammation RR 1.04 0.66 1.42 Very_low 3
  Lou_MWC_2023 Adiponectin Breast_Cancer Inflammation RR 0.76 0.61 0.91 Moderate 8
  Lou_MWC_2023 Leptin Breast_Cancer Inflammation RR 0.9 0.63 1.18 Low 8",
  header = TRUE
)

# 定义关联方向判定函数
get_direction <- function(es_measure, lower, upper, effect_type) {
  if (effect_type == "SMD") {
    if (es_measure > 0 & lower > 0 & upper > 0) {
      return("Positive")
    } else if (es_measure < 0 & lower < 0 & upper < 0) {
      return("Negative")
    } else {
      return("Null")
    }
  } else if (effect_type == "RR") {
    if (es_measure > 1 & lower > 1 & upper > 1) {
      return("Positive")
    } else if (es_measure < 1 & lower < 1 & upper < 1) {
      return("Negative")
    } else {
      return("Null")
    }
  }
}

# 计算每条路径的关联方向
data$direction <- with(data, mapply(get_direction, exposure_int_es, 
                                    exposure_int_lower, exposure_it_higher, 
                                    effect_type = exp_in_es_measure))

可视化实现代码 #

使用tidygraph构建图数据，ggraph绘制路径图：

# 加载所需包
library(tidygraph)
library(ggraph)
library(dplyr)

# 1. 整理节点数据：提取所有节点，设置标注
nodes <- data %>%
  select(Exposure, Outcome) %>%
  pivot_longer(cols = everything(), values_to = "name") %>%
  distinct(name) %>%
  mutate(label = case_when(
    name == "Physical_activity" ~ "Exposure",
    name == "Breast_Cancer" ~ "Outcome",
    TRUE ~ name
  ))

# 2. 整理边数据：设置箭头属性，生成标注文本
edges <- data %>%
  mutate(
    # 计算箭头粗细：取效应量绝对值，做标准化处理
    edge_width = abs(exposure_int_es) / max(abs(exposure_int_es)) * 3,
    # 生成效应量+95%CI标注文本
    edge_label = paste0(
      ifelse(exp_in_es_measure == "SMD", round(exposure_int_es, 2), exposure_int_es),
      " (", round(exposure_int_lower, 2), "–", round(exposure_it_higher, 2), ")"
    )
  ) %>%
  select(from = Exposure, to = Outcome, direction, edge_width, edge_label)

# 3. 构建图对象
graph <- tbl_graph(nodes = nodes, edges = edges, directed = TRUE)

# 4. 绘制路径图
ggraph(graph, layout = "linear", circular = FALSE) +
  # 绘制箭头
  geom_edge_link(aes(color = direction, width = edge_width), 
                 arrow = arrow(type = "closed", length = unit(4, "mm")),
                 end_cap = circle(5, "mm")) +
  # 绘制节点
  geom_node_label(aes(label = label), size = 4, fill = "white", color = "black") +
  # 设置箭头颜色映射
  scale_color_manual(values = c("Positive" = "red", "Negative" = "blue", "Null" = "gray")) +
  # 设置箭头粗细范围
  scale_width_continuous(range = c(0.5, 3)) +
  # 调整主题：去掉坐标轴和背景
  theme_graph() +
  labs(color = "关联方向") +
  # 调整布局，让节点水平排列
  coord_cartesian(xlim = c(-0.5, nrow(nodes)-0.5))

代码说明 #

• 节点处理：将暴露和结局节点替换为要求的标注，中介节点保留原名称
• 箭头处理：用效应量绝对值的标准化值控制粗细，根据direction变量设置颜色（红=正相关，蓝=负相关，灰=无关联）
• 标注处理：自动生成包含效应量和95%CI的文本，添加在箭头上
• 布局：采用线性水平布局，清晰展示暴露→中介→结局的路径关系

内容的提问来源于stack exchange，提问作者user115916