ggalluvial|炫酷桑基图（Sankey），你也可以秀

参考：https://mp.weixin.qq.com/s?__biz=MzIyNDI1MzgzOQ==&mid=2650394249&idx=1&sn=d67822590e8eceb9b4f544b4e2946d37&chksm=f01cac69c76b257f57067e8891379183e15ae733118cfa2230bd4e9c3a3539803b4b94abe144&scene=21#wechat_redirect

桑基图（Sankey diagram），是一种特定类型的流程图，图中延伸的分支的宽度对应数据流量的大小，通常应用于能源、材料成分、金融等数据的可视化分析。

因1898年Matthew Henry Phineas Riall Sankey绘制的“蒸汽机的能源效率图”而闻名，此后便以其名字命名为“桑基图”。

载入R包，数据

本文使用TCGA数据集中的LIHC的临床数据进行展示，大家可以根据数据格式处理自己的临床数据。也可后台回复“R-桑基图”获得示例数据以及R代码。

#install.packages("ggalluvial")
library(ggalluvial)
library(ggplot2)
library(dplyr)
#读入LIHC临床数据
LIHC <- read.csv("TCGA_lihc.csv",header=TRUE)
#展示数据情况
head(LIHC)
   PATIENT_ID    AGE    SEX AJCC_PATHOLOGIC_TUMOR_STAGE OS_STATUS
1 TCGA-XR-A8TE less50   Male                   STAGE III    LIVING
2 TCGA-5R-AA1D less50 Female                   STAGE III    LIVING
3 TCGA-DD-A1EC less50 Female                     STAGE I    LIVING
4 TCGA-ED-A7PY less50 Female                    STAGE II    LIVING
5 TCGA-RC-A6M5 less50 Female                    STAGE IV    LIVING
6 TCGA-DD-A1EH less50   Male                   STAGE III    LIVING

summary(LIHC)

桑基图的数据结构需要节点，权重等信息，ggalluvial 的输入数据可以是长数据亦可以是宽数据。

绘制桑基图

1 宽数据示例

    对临床数据进行简单的处理，得到后四个变量的频数，整理成宽数据：以下处理过程可参考数据处理|R-dplyr，数据处理|数据框重铸

#分组计算频数
LIHCData <- group_by(data,AGE,SEX,AJCC_PATHOLOGIC_TUMOR_STAGE,OS_STATUS) %>% summarise(., count = n())
#查看宽数据格式
head(LIHCData)
 AGE    SEX    AJCC_PATHOLOGIC_TUMOR_STAGE OS_STATUS count
 <fct>  <fct>  <fct>                       <fct>     <int>
1 50to70 Female STAGE I                     DECEASED     11
2 50to70 Female STAGE I                     LIVING       16
3 50to70 Female STAGE II                    DECEASED      3
4 50to70 Female STAGE II                    LIVING       11
5 50to70 Female STAGE III                   DECEASED      8
6 50to70 Female STAGE III                   LIVING        9

绘制桑基图

ggplot(as.data.frame(LIHCData),
      aes(axis1 = AJCC_PATHOLOGIC_TUMOR_STAGE, axis2 = SEX, axis3 = AGE,
          y= count)) +
 scale_x_discrete(limits = c("AJCC_STAGE", "SEX", "AGE"), expand = c(.1, .05)) +
 geom_alluvium(aes(fill = OS_STATUS)) +
 geom_stratum() + geom_text(stat = "stratum", label.strata = TRUE) +
 theme_minimal() +
 ggtitle("Patients in the TCGA-LIHC cohort",
         "stratified by demographics and survival")

• axis参数设置待展示的节点信息（柱子）；
• geom_alluvium参数设置组间面积连接，此处按生存状态分组；

2 长数据示例

ggplot2通常处理的都是长表格模式，使用to_lodes_form函数即可转换

#to_lodes_form生成alluvium和stratum列，主分组位于key列中
LIHC_long <- to_lodes_form(data.frame(LIHCData),
                             key = "Demographic",
                             axes = 1:3)
head(LIHC_long)



 OS_STATUS count alluvium Demographic stratum
1  DECEASED    11        1         AGE  50to70
2    LIVING    16        2         AGE  50to70
3  DECEASED     3        3         AGE  50to70
4    LIVING    11        4         AGE  50to70
5  DECEASED     8        5         AGE  50to70
6    LIVING     9        6         AGE  50to70

# 绘制桑基图
ggplot(data = LIHC_long,
      aes(x = Demographic, stratum = stratum, alluvium = alluvium,
          y = count, label = stratum)) +
 geom_alluvium(aes(fill = OS_STATUS)) +
 geom_stratum() + geom_text(stat = "stratum") +
 theme_minimal() +
 ggtitle("Patients in the TCGA-LIHC cohort",
         "stratified by demographics and survival")

3 状态变化的趋势

    vaccinations为R包内置数据集，可展示同一subject在不同survey状态下的response情况。

data(vaccinations)
levels(vaccinations$response) <- rev(levels(vaccinations$response))
ggplot(vaccinations,
      aes(x = survey, stratum = response, alluvium = subject,
          y = freq,
          fill = response, label = response)) +
 scale_x_discrete(expand = c(.1, .1)) +
 geom_flow() +
 geom_stratum(alpha = .5) +
 geom_text(stat = "stratum", size = 3) +
 theme(legend.position = "none") +
 ggtitle("vaccination survey responses at three points in time")

4 更多细节

vignette(topic = "ggalluvial", package = "ggalluvial")

 以上就是如何使用R-ggalluvial包绘制桑基图的简单介绍，可以自己动手展示了 。