funkyheatmap |临床+组学+分组数据可视化“神器”，时髦的热图

临床数据一般是使用图表汇总Table1的方式进行展示，例如R|tableone 快速绘制文章“表一”-基线特征三线表 或者 gtsummary|巧合-绘制多种数据汇总表“神器” 。

今天介绍一个可视化展示方式，funkyheatmap-R包 ， 可以为基准数据生成热图式可视化的函数，可以使用列和行的注释对其进行微调 。效果如下

一 载入R包，数据 

首先安装funkyheatmap 包，

1）先使用mtcars 数据绘制

#devtools::install_github("dynverse/dynbenchmark/package")install.packages("funkyheatmap")library(funkyheatmap)library(dplyr, warn.conflicts = FALSE)library(tibble, warn.conflicts = FALSE)library(tidyverse)data("mtcars")funky_heatmap(mtcars)

使用mtcars类似的数据进行绘制即可，默认情况下绘制每一列的信息，后面就是参数修改以期达到封面图的效果。

下面使用dynbenchmark_data的数据进行详细的调试 绘制。

2）载入dynbenchmark_data数据（2019 NBT ：A comparison of single-cell trajectory inference methods ）

data("dynbenchmark_data")data <- dynbenchmark_datahead(data)

二 绘制funky heatmap

想达到封面图的效果，需要一系列的设置。

1 , 设置row_info 和 row_group

row_info选择待展示的行（此处即为id列各种方法），此处为全部展示，可是使用filter筛选想展示的行；

row_groups是行group（此处为各种方法的类型），对应下图的红框部分

row_info <-  data %>%  select(group, id)row_groups <-  data %>%  transmute(    group,    Group = case_when(      group == "cycle" ~ "Cyclic methods",      TRUE ~ paste0(stringr::str_to_title(group), " methods")    )  ) %>%  unique()head(row_info)# A tibble: 6 × 2#  group id                 #  <fct> <chr>              #1 graph paga               #2 graph raceid_stemid      #3 graph slicer             #4 tree  slingshot          #5 tree  paga_tree          #6 tree  projected_slingshothead(row_groups)# A tibble: 6 × 2#  group          Group                 #  <fct>          <chr>                 #1 graph          Graph methods         #2 tree           Tree methods          #3 multifurcation Multifurcation methods#4 bifurcation    Bifurcation methods   #5 linear         Linear methods        #6 cycle          Cyclic methods

2 ，设置column_info

设置列的信息，定义需要展示的列以及对应的一些属性信息

column_info <- tribble( # tribble_start  ~group,                   ~id,                                            ~name,                      ~geom,        ~palette,         ~options,  "method_characteristic",  "method_name",                                  "",                         "text",       NA,               list(hjust = 0, width = 6),  "method_characteristic",  "method_platform",                              "Platform",                 "text",       NA,               list(width = 2),  "method_characteristic",  "method_topology_inference",                    "Topology inference",       "text",       NA,               list(width = 2),  "score_overall",          "summary_overall_overall",                      "Overall",                  "bar",        "overall",        list(width = 4, legend = FALSE),  "score_overall",          "benchmark_overall_overall",                    "Accuracy",                 "bar",        "benchmark",      list(width = 4, legend = FALSE),  "score_overall",          "qc_overall_overall",                           "Usability",                "bar",        "qc",             list(width = 4, legend = FALSE),  "score_overall",          "control_label",                                "",                         "text",       NA,               list(overlay = TRUE),  "benchmark_metric",       "benchmark_overall_norm_him",                   "Topology",                 "funkyrect",  "benchmark",      lst(),  "benchmark_metric",       "benchmark_overall_norm_F1_branches",           "Branch assignment",        "funkyrect",  "benchmark",      lst(),  "benchmark_metric",       "benchmark_overall_norm_correlation",           "Cell positions",           "funkyrect",  "benchmark",      lst(),  "benchmark_metric",       "benchmark_overall_norm_featureimp_wcor",       "Features",                 "funkyrect",  "benchmark",      lst(),  "benchmark_source",       "benchmark_source_real_gold",                   "Gold",                     "funkyrect",  "benchmark",      lst(),  "benchmark_source",       "benchmark_source_real_silver",                 "Silver",                   "funkyrect",  "benchmark",      lst(),  "benchmark_source",       "benchmark_source_synthetic_dyngen",            "dyngen",                   "funkyrect",  "benchmark",      lst(),  "benchmark_source",       "benchmark_source_synthetic_dyntoy",            "dyntoy",                   "funkyrect",  "benchmark",      lst(),  "benchmark_source",       "benchmark_source_synthetic_prosstt",           "PROSSTT",                  "funkyrect",  "benchmark",      lst(),  "benchmark_source",       "benchmark_source_synthetic_splatter",          "Splatter",                 "funkyrect",  "benchmark",      lst(),  "benchmark_execution",    "benchmark_overall_pct_errored_str",            "% Errored",                "text",       NA,               lst(hjust = 1),  "benchmark_execution",    "benchmark_overall_error_reasons",              "Reason",                   "pie",        "error_reasons",  lst(),  "scaling_predtime",       "scaling_pred_scoretime_cells1m_features100",   "1m \u00D7 100",            "rect",       "scaling",        lst(scale = FALSE),  "scaling_predtime",       "scaling_pred_scoretime_cells1m_features100",   "",                         "text",       "white6black4",   lst(label = "scaling_pred_timestr_cells1m_features100", overlay = TRUE, size = 3, scale = FALSE),  "scaling_predtime",       "scaling_pred_scoretime_cells100k_features1k",  "100k \u00D7 1k",           "rect",       "scaling",        lst(scale = FALSE),  "scaling_predtime",       "scaling_pred_scoretime_cells100k_features1k",  "",                         "text",       "white6black4",   lst(label = "scaling_pred_timestr_cells100k_features1k", overlay = TRUE, size = 3, scale = FALSE),  "scaling_predtime",       "scaling_pred_scoretime_cells10k_features10k",  "10k \u00D7 10k",           "rect",       "scaling",        lst(scale = FALSE),  "scaling_predtime",       "scaling_pred_scoretime_cells10k_features10k",  "",                         "text",       "white6black4",   lst(label = "scaling_pred_timestr_cells10k_features10k", overlay = TRUE, size = 3, scale = FALSE),  "scaling_predtime",       "scaling_pred_scoretime_cells1k_features100k",  "1k \u00D7 100k",           "rect",       "scaling",        lst(scale = FALSE),  "scaling_predtime",       "scaling_pred_scoretime_cells1k_features100k",  "",                         "text",       "white6black4",   lst(label = "scaling_pred_timestr_cells1k_features100k", overlay = TRUE, size = 3, scale = FALSE),  "scaling_predtime",       "scaling_pred_scoretime_cells100_features1m",   "100 \u00D7 1m",            "rect",       "scaling",        lst(scale = FALSE),  "scaling_predtime",       "scaling_pred_scoretime_cells100_features1m",   "",                         "text",       "white6black4",   lst(label = "scaling_pred_timestr_cells100_features1m", overlay = TRUE, size = 3, scale = FALSE),  "scaling_predtime",       "benchmark_overall_time_predcor_str",           "Cor. pred. vs. real",      "text",       NA,               lst(size = 3),  "stability",              "stability_him",                                "Topology",                 "funkyrect",  "stability",      lst(),  "stability",              "stability_F1_branches",                        "Branch assignment",        "funkyrect",  "stability",      lst(),  "stability",              "stability_correlation",                        "Cell positions",           "funkyrect",  "stability",      lst(),  "stability",              "stability_featureimp_wcor",                    "Features",                 "funkyrect",  "stability",      lst(),  "qc_category",            "qc_cat_availability",                          "Availability",             "funkyrect",  "qc",             lst(),  "qc_category",            "qc_cat_behaviour",                             "Behaviour",                "funkyrect",  "qc",             lst(),  "qc_category",            "qc_cat_code_assurance",                        "Code assurance",           "funkyrect",  "qc",             lst(),  "qc_category",            "qc_cat_code_quality",                          "Code quality",             "funkyrect",  "qc",             lst(),  "qc_category",            "qc_cat_documentation",                         "Documentation",            "funkyrect",  "qc",             lst(),  "qc_category",            "qc_cat_paper",                                 "Paper",                    "funkyrect",  "qc",             lst(),  "qc_category",            "control_label",                                "",                         "text",       NA,               list(overlay = TRUE, width = -6))

主要有以下几列：

group：列的分组；

id：data中的列名字；

name：图中展示的名字；

geom：集合图形，展示方式 （如果有多个属性（rect 和 text）需要展示则分为多行，如scaling_predtime）；

palette：调色板信息；

3，设置column_groups

将上述column_info的列，根据对应的group ，设置group的 “Category”和大一级的 Experiment 信息。

column_groups <- tribble(  ~Experiment,    ~Category,                                      ~group,                   ~palette,  "Method",       "\n",                                           "method_characteristic",  "overall",  "Summary",      "Aggregated scores per experiment",             "score_overall",          "overall",  "Accuracy",     "Per metric",                                   "benchmark_metric",       "benchmark",  "Accuracy",     "Per dataset source",                           "benchmark_source",       "benchmark",  "Accuracy",     "Errors",                                       "benchmark_execution",    "benchmark",  "Scalability",  "Predicted time\n(#cells \u00D7 #features)",    "scaling_predtime",       "scaling",  "Stability",    "Similarity\nbetween runs",                     "stability",              "stability",  "Usability",    "Quality of\nsoftware and paper",               "qc_category",            "qc")

Experiment：group的Experiment信息（下图红框）

Category：group的Category信息（下图绿框）

group：列的分组（同column_info中的group）

palette：group使用何种palette 

4，设置 palettes

设置不同palette的颜色

error_reasons <- tibble(  name = c("pct_memory_limit", "pct_time_limit", "pct_execution_error", "pct_method_error"),  label = c("Memory limit exceeded", "Time limit exceeded", "Execution error", "Method error"),  colour = RColorBrewer::brewer.pal(length(name), "Set3"))
palettes <- tribble(  ~palette,             ~colours,  "overall",            grDevices::colorRampPalette(rev(RColorBrewer::brewer.pal(9, "Greys")[-1]))(101),  "benchmark",          grDevices::colorRampPalette(rev(RColorBrewer::brewer.pal(9, "Blues") %>% c("#011636")))(101),  "scaling",            grDevices::colorRampPalette(rev(RColorBrewer::brewer.pal(9, "Reds")[-8:-9]))(101),  "stability",          grDevices::colorRampPalette(rev(RColorBrewer::brewer.pal(9, "YlOrBr")[-7:-9]))(101),  "qc",                 grDevices::colorRampPalette(rev(RColorBrewer::brewer.pal(9, "Greens")[-1] %>% c("#00250f")))(101),  "error_reasons",      error_reasons %>% select(label, colour) %>% deframe(),  "white6black4",       c(rep("white", 3), rep("black", 7)))

5，绘制funky heatmap

完成上述设置后终于可以绘制funky heatmap了

g <- funky_heatmap(  data = data,  column_info = column_info,  column_groups = column_groups,  row_info = row_info,  row_groups = row_groups,  palettes = palettes,  col_annot_offset = 3.2)g#保存输出#ggsave("path_to_plot.pdf", g, device = cairo_pdf, width = g$width, height = g$height)

OK，这样每个患者（id）的临床信息，组学信息，分组信息，就都可以可视化展示了！

参考资料：

http://funkyheatmap.dynverse.org/

http://funkyheatmap.dynverse.org/articles/dynbenchmark.html