求助：使用dendextend的color_branches后热图分支颜色异常

尝试使用dendextend包的color_branches()函数为ComplexHeatmap绘制的热图树状图分支上色时，出现部分叶节点呈现混合颜色的异常。相关代码如下：

column_dend = hclust(dist(t(data)))
row_dend = hclust(dist((data)))
row_dend = color_branches(row_dend, k = 2)

ht<-Heatmap(as.matrix(data),
            name = "legend",
            cluster_rows = row_dend,
            cluster_columns = column_dend,
            na_col = 'black',
            column_names_rot = 45,
            col = circlize::colorRamp2(c(-1, 0, 1), c("#56B4E9", "#FFFFFF", "#FF7400")),
            show_row_names = F,
            show_column_names = F,
            use_raster = F,
            heatmap_width = unit(1.5, "cm")*ncol(data),
            column_names_gp = grid::gpar(fontsize = 13),
            row_names_gp = grid::gpar(fontsize = 1))

ht = draw(ht)

• 核心原因：dendextend::color_branches()返回的是dendrogram对象，ComplexHeatmap的cluster_rows参数虽支持传入该类型，但二者在树状图渲染的细节处理上存在兼容性问题——color_branches()会给分支添加颜色属性，而ComplexHeatmap在渲染叶节点附近的短分支时，会出现颜色叠加绘制的情况，最终导致叶节点呈现混合色。

• 解决办法：

  1. 改用ComplexHeatmap原生分支上色方式，无需依赖dendextend：
     绘制热图后，通过decorate_row_dend()函数为行树状图分支上色，示例代码如下：

     column_dend = hclust(dist(t(data)))
     row_dend = hclust(dist(data))
     
     ht<-Heatmap(as.matrix(data),
                 name = "legend",
                 cluster_rows = row_dend,
                 cluster_columns = column_dend,
                 na_col = 'black',
                 column_names_rot = 45,
                 col = circlize::colorRamp2(c(-1, 0, 1), c("#56B4E9", "#FFFFFF", "#FF7400")),
                 show_row_names = F,
                 show_column_names = F,
                 use_raster = F,
                 heatmap_width = unit(1.5, "cm")*ncol(data),
                 column_names_gp = grid::gpar(fontsize = 13),
                 row_names_gp = grid::gpar(fontsize = 1))
     
     ht = draw(ht)
     # 为行树状图分支上色
     decorate_row_dend("ht", {
       dend = current_dendrogram()
       dend = color_branches(dend, k = 2)
       grid::grid.draw(dend)
     })
     

  2. 若坚持使用dendextend处理结果，可调整树状图的叶节点悬挂长度，避免短分支导致的颜色叠加：

     row_dend = hclust(dist(data))
     row_dend = as.dendrogram(row_dend)
     row_dend = color_branches(row_dend, k = 2)
     # 调整叶节点悬挂长度，数值可按需调整
     row_dend = hang.dendrogram(row_dend, hang = 0.1)
     

内容的提问来源于stack exchange，提问作者Romina Appierdo