基于GOA与depthcenter排序ggplot2气泡图X轴Sample变量

我有一个超10000行的数据集，已经用ggplot2绘制了气泡图，但需要对坐标轴做自定义排序。我知道可以用factor()结合levels列表来排序变量，但变量数量太多（有数百个基因注释变量），手动编写levels完全不现实。

我之前使用factor()的示例：

primary_order_list4 <- c("RHM","TS","SCS","LIL","STN","STS") 
fenorm.gene.count$Pond <- factor(fenorm.gene.count$Pond,levels=primary_order_list4)

X轴排序需求 #

Sample列格式为「地点+年份+深度」（例如SCS19-3），排序规则：

• 先按对应池塘的GOA值升序排序（比如GOA为6.4的RHM要排在GOA为18.7的SCS之前）
• 同一池塘内，再按depthcenter从浅（0.5）到深（19.5）排序

我尝试过先按depthcenter和GOA升序排序数据框，但ggplot绘制出的X轴仍为字母顺序：

fenorm.gene.count <- fenorm.gene.count[order(fenorm.gene.count$depthcenter),]
ugh <- fenorm.gene.count[order(fenorm.gene.count$GOA),]
ughplot <- ggplot(fenorm.gene.count, aes(x=Sample,y=HMM,fill=Pond)) + 
  geom_point(aes(size=ifelse(total_count==0, NA, total_count)))

另外Y轴（HMM列，基因注释）也需要类似的自定义排序逻辑。

数据集示例 #

dput(head(fenorm.gene.count))
structure(list(Project = c("Ga0598240", "Ga0598240", "Ga0598240", 
"Ga0598240", "Ga0598240", "Ga0598240"), category = c("iron_reduction", 
"iron_reduction", "iron_oxidation", "iron_oxidation", "iron_reduction", 
"iron_oxidation"), HMM = c("MtrC_TIGR03507", 
"OmcS", "OmcF", "Cyc2_repCluster2", "PF01032-FecCD-YfhA-FpvE-YfeCD_transport_family", 
"PvuCD-FhuB-CbrBC-FeuB-CbrC-YfhA"), total_count = c(10L, 
1L, 17L, 55L, 15L, 1L), norm = c(0.00136836343732895, 0.000136836343732895, 
0.00232621784345922, 0.00752599890530925, 0.00205254515599343, 
0.000136836343732895), Sample = c("LIL19-1", "LIL19-1", "LIL19-1", 
"LIL19-1", "LIL19-1", "LIL19-1"), Portal = c(3300064049, 3300064049, 
3300064049, 3300064049, 3300064049, 3300064049), G2 = c("N", 
"N", "N", "N", "N", "N"), Pond = c("LIL", "LIL", "LIL", "LIL", 
"LIL", "LIL"), year = c(2019L, 2019L, 2019L, 2019L, 2019L, 2019L
), depthcenter = c(0.5, 0.5, 0.5, 0.5, 0.5, 0.5), toplow = c("top", 
"top", "top", "top", "top", "top"), GOA = c(19.4, 19.4, 19.4, 
19.4, 19.4, 19.4), elevation = c(8.2, 8.2, 8.2, 8.2, 8.2, 8.2
)), row.names = 65:70, class = "data.frame")

1. X轴（Sample）排序 #

核心思路是先基于GOA和depthcenter自动生成排序规则，再将Sample转为带对应levels的因子，无需手动编写levels：

# 提取每个Sample对应的唯一GOA和depthcenter值（同一Sample的这两个值一致）
sample_order_df <- fenorm.gene.count %>%
  distinct(Sample, Pond, GOA, depthcenter) %>%
  arrange(GOA, depthcenter)  # 按GOA升序、再按depthcenter升序排列

# 将Sample转为因子，levels取排序后的Sample值
fenorm.gene.count$Sample <- factor(fenorm.gene.count$Sample, levels = sample_order_df$Sample)

2. Y轴（HMM）排序 #

如果Y轴需要按分组或统计指标排序（比如按category分组后，再按total_count均值降序），用同样逻辑自动生成levels：

# 按category分组，计算每个HMM的total_count均值并排序
hmm_order_df <- fenorm.gene.count %>%
  group_by(category, HMM) %>%
  summarise(mean_count = mean(total_count), .groups = "drop") %>%
  arrange(category, desc(mean_count))  # 先按category排序，再按均值降序

# 将HMM转为因子，levels取排序后的HMM值
fenorm.gene.count$HMM <- factor(fenorm.gene.count$HMM, levels = hmm_order_df$HMM)

3. 重新绘制气泡图 #

现在绘图时，坐标轴会自动按自定义规则排序：

ughplot <- ggplot(fenorm.gene.count, aes(x=Sample, y=HMM, fill=Pond)) + 
  geom_point(aes(size=ifelse(total_count==0, NA, total_count))) +
  # 若X轴标签拥挤，可添加旋转优化显示
  theme(axis.text.x = element_text(angle = 45, hjust = 1))

关键说明 #

• 仅排序数据框不会改变ggplot的坐标轴顺序，因为ggplot默认将字符型变量视为无序因子，按字母排序。必须将变量转为带自定义levels的因子才能生效。
• 用dplyr的distinct()和arrange()可自动生成排序后的levels列表，完美适配海量变量场景，无需手动输入。

内容的提问来源于stack exchange，提问作者Geomicro