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_data
head(data)
二 绘制funky heatmap
想达到封面图的效果,需要一系列的设置。
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_slingshot
head(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
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;
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
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))
)
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