Citation

Mangiola et al., (2020). tidyHeatmap: an R package for modular heatmap production based on tidy principles. Journal of Open Source Software, 5(52), 2472, https://doi.org/10.21105/joss.02472

Please have a look also to

• tidygate for adding custom gate information to your tibble
• tidySingleCellExperiment for tidy manipulation of Seurat objects
• tidyseurat for tidy manipulation of Seurat objects
• tidybulk for tidy high-level data analysis and manipulation
• tidySummarizedExperiment for heatmaps produced with tidy principles

website: stemangiola.github.io/tidyHeatmap

tidyHeatmap is a package that introduces tidy principles to the creation of information-rich heatmaps. This package uses ComplexHeatmap as graphical engine.

Advantages:

• Modular annotation with just specifying column names
• Custom grouping of rows is easy to specify providing a grouped tbl. For example df |> group_by(...)
• Labels size adjusted by row and column total number
• Default use of Brewer and Viridis palettes

Functions/utilities available

Installation

To install the most up-to-date version

devtools::install_github("stemangiola/tidyHeatmap")

To install the most stable version (however please keep in mind that this package is under a maturing lifecycle stage)

install.packages("tidyHeatmap")

Contribution

If you want to contribute to the software, report issues or problems with the software or seek support please open an issue here

Input data frame

The heatmaps visualise a multi-element, multi-feature dataset, annotated with independent variables. Each observation is a element-feature pair (e.g., person-physical characteristics).

Let’s transform the mtcars dataset into a tidy “element-feature-independent variables” data frame. Where the independent variables in this case are ‘hp’ and ‘vs’.

mtcars_tidy <- 
    mtcars |> 
    as_tibble(rownames="Car name") |> 
    
    # Scale
    mutate_at(vars(-`Car name`, -hp, -vs), scale) |>
    
    # tidyfy
    pivot_longer(cols = -c(`Car name`, hp, vs), names_to = "Property", values_to = "Value")

mtcars_tidy

## # A tibble: 288 × 5
##    `Car name`       hp    vs Property Value[,1]
##    <chr>         <dbl> <dbl> <chr>        <dbl>
##  1 Mazda RX4       110     0 mpg          0.151
##  2 Mazda RX4       110     0 cyl         -0.105
##  3 Mazda RX4       110     0 disp        -0.571
##  4 Mazda RX4       110     0 drat         0.568
##  5 Mazda RX4       110     0 wt          -0.610
##  6 Mazda RX4       110     0 qsec        -0.777
##  7 Mazda RX4       110     0 am           1.19 
##  8 Mazda RX4       110     0 gear         0.424
##  9 Mazda RX4       110     0 carb         0.735
## 10 Mazda RX4 Wag   110     0 mpg          0.151
## # … with 278 more rows

Plotting

For plotting, you simply pipe the input data frame into heatmap, specifying:

• The rows, cols relative column names (mandatory)
• The value column name (mandatory)
• The annotations column name(s)

mtcars

mtcars_heatmap <- 
    mtcars_tidy |> 
    heatmap(`Car name`, Property, Value ) |>
    add_tile(hp)

## tidyHeatmap says: (once per session) from release 1.2.3 the grouping labels have white background by default. To add color for one-ay grouping specify palette_grouping = list(c("red", "blue"))

mtcars_heatmap

Saving

mtcars_heatmap |> save_pdf("mtcars_heatmap.pdf")

Grouping and splitting

We can easily group the data (one group per dimension maximum, at the moment only the vertical dimension is supported) with dplyr, and the heatmap will be grouped accordingly

# Make up more groupings
mtcars_tidy_groupings = 
    mtcars_tidy |>
    mutate(property_group = if_else(Property %in% c("cyl", "disp"), "Engine", "Other"))

mtcars_tidy_groupings |> 
    group_by(vs, property_group) |>
    heatmap(`Car name`, Property, Value ) |>
    add_tile(hp)

We can provide colour palettes to groupings

mtcars_tidy_groupings |> 
    group_by(vs, property_group) |>
    heatmap(
        `Car name`, Property, Value ,
        palette_grouping = list(
            
            # For first grouping (vs)
            c("#66C2A5", "#FC8D62"), 
            
            # For second grouping (property_group)
            c("#b58b4c", "#74a6aa")
        )
    ) |>
    add_tile(hp)

We can split based on the cladogram

mtcars_tidy |> 
    heatmap(`Car name`, Property, Value ) |>
    split_rows(2) |>
    split_columns(2)

We can split on kmean clustering (using ComplexHeatmap options, it is stochastic)

mtcars_tidy |> 
    heatmap(
        `Car name`, Property, Value ,
        row_km = 2,
        column_km = 2
    ) 

Custom palettes

We can easily use custom palette, using strings, hexadecimal color character vector,

mtcars_tidy |> 
    heatmap(
        `Car name`, 
        Property, 
        Value,
        palette_value = c("red", "white", "blue")
    )

A better-looking blue-to-red palette

mtcars_tidy |> 
    heatmap(
        `Car name`, 
        Property, 
        Value,
        palette_value = circlize::colorRamp2(
            seq(-2, 2, length.out = 11), 
            RColorBrewer::brewer.pal(11, "RdBu")
        )
    )

Or a grid::colorRamp2 function for higher flexibility

mtcars_tidy |> 
    heatmap(
        `Car name`, 
        Property, 
        Value,
        palette_value = circlize::colorRamp2(c(-2, -1, 0, 1, 2), viridis::magma(5))
    )

Multiple groupings and annotations

tidyHeatmap::pasilla |>
    group_by(location, type) |>
    heatmap(
        .column = sample,
        .row = symbol,
        .value = `count normalised adjusted`
    ) |>
    add_tile(condition) |>
    add_tile(activation)

Remove legends, adding aesthetics to annotations in a modular fashion, using ComplexHeatmap arguments

tidyHeatmap::pasilla |>
    group_by(location, type) |>
    heatmap(
        .column = sample,
        .row = symbol,
        .value = `count normalised adjusted`,
        show_heatmap_legend = FALSE
    ) |>
    add_tile(condition, show_legend = FALSE) |>
    add_tile(activation, show_legend = FALSE)

Annotation types

“tile”, “point”, “bar” and “line” are available

# Create some more data points
pasilla_plus <- 
    tidyHeatmap::pasilla |>
    dplyr::mutate(act = activation) |> 
    tidyr::nest(data = -sample) |>
    dplyr::mutate(size = rnorm(n(), 4,0.5)) |>
    dplyr::mutate(age = runif(n(), 50, 200)) |>
    tidyr::unnest(data) 

# Plot
pasilla_plus |>
    heatmap(
        .column = sample,
        .row = symbol,
        .value = `count normalised adjusted`
    ) |>
    add_tile(condition) |>
    add_point(activation) |>
    add_tile(act) |>
    add_bar(size) |>
    add_line(age)

Annotation size

We can customise annotation sizes using the grid::unit(), and the size of their names using in-built ComplexHeatmap arguments

pasilla_plus |>
    heatmap(
        .column = sample,
        .row = symbol,
        .value = `count normalised adjusted`
    ) |>
    add_tile(condition, size = unit(0.3, "cm"), annotation_name_gp= gpar(fontsize = 8)) |>
    add_point(activation, size = unit(0.3, "cm"),   annotation_name_gp= gpar(fontsize = 8)) |>
    add_tile(act, size = unit(0.3, "cm"),   annotation_name_gp= gpar(fontsize = 8)) |>
    add_bar(size, size = unit(0.3, "cm"),   annotation_name_gp= gpar(fontsize = 8)) |>
    add_line(age, size = unit(0.3, "cm"),   annotation_name_gp= gpar(fontsize = 8))

Layer symbol

Add a layer on top of the heatmap

tidyHeatmap::pasilla |>
    
    # filter
    filter(symbol %in% head(unique(tidyHeatmap::pasilla$symbol), n = 10)) |>
    
    heatmap(
        .column = sample,
        .row = symbol,
        .value = `count normalised adjusted`
    ) |> 
    layer_point(
        `count normalised adjusted log` > 6 & sample == "untreated3" 
    )

ComplexHeatmap further styling

mtcars_tidy |> 
    heatmap(
        `Car name`, Property, Value, 
        rect_gp = grid::gpar(col = "#161616", lwd = 0.5)
    ) 

mtcars_tidy |> 
    heatmap(
        `Car name`, Property, Value, 
        cluster_rows = FALSE
    ) 

library(forcats)

## Warning: package 'forcats' was built under R version 4.1.2

mtcars_tidy |> 
    mutate(`Car name` = fct_reorder(`Car name`, `Car name`, .desc = TRUE)) %>% 
    heatmap(
        `Car name`, Property, Value, 
        cluster_rows = FALSE
    ) 

mtcars_tidy |> 
    mutate(`Car name` = fct_reorder(`Car name`, `Car name`, .desc = TRUE)) %>% 
    heatmap(
        `Car name`, Property, Value, 
        column_dend_height = unit(0.2, "cm"), 
        row_dend_width = unit(0.2, "cm")
    ) 

mtcars_tidy |> 
    mutate(`Car name` = fct_reorder(`Car name`, `Car name`, .desc = TRUE)) %>% 
    heatmap(
        `Car name`, Property, Value, 
        row_names_gp = gpar(fontsize = 7),
        column_names_gp = gpar(fontsize = 7),
        column_title_gp = gpar(fontsize = 7),
        row_title_gp = gpar(fontsize = 7)
    ) 

library(ggplot2)
library(patchwork)

p_heatmap =
    mtcars_tidy |> 
    heatmap(
        `Car name`, Property, Value, 
            show_heatmap_legend = FALSE,
        row_names_gp = gpar(fontsize = 7)
    ) 

p_ggplot = tibble(value = 1:10) %>% ggplot(aes(value)) + geom_density()

wrap_heatmap(p_heatmap) + 
    p_ggplot +
    wrap_heatmap(p_heatmap) + 
    plot_layout(width = c(1, 0.3, 1))

Session Info

sessionInfo()

## R version 4.1.0 (2021-05-18)
## Platform: x86_64-pc-linux-gnu (64-bit)
## Running under: CentOS Linux 7 (Core)
## 
## Matrix products: default
## BLAS:   /stornext/System/data/apps/R/R-4.1.0/lib64/R/lib/libRblas.so
## LAPACK: /stornext/System/data/apps/R/R-4.1.0/lib64/R/lib/libRlapack.so
## 
## locale:
##  [1] LC_CTYPE=en_US.UTF-8       LC_NUMERIC=C              
##  [3] LC_TIME=en_US.UTF-8        LC_COLLATE=C              
##  [5] LC_MONETARY=en_US.UTF-8    LC_MESSAGES=en_US.UTF-8   
##  [7] LC_PAPER=en_US.UTF-8       LC_NAME=C                 
##  [9] LC_ADDRESS=C               LC_TELEPHONE=C            
## [11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C       
## 
## attached base packages:
## [1] grid      stats     graphics  grDevices utils     datasets  methods  
## [8] base     
## 
## other attached packages:
## [1] patchwork_1.1.1   ggplot2_3.3.5     forcats_0.5.1     tidyHeatmap_1.6.0
## [5] tidyr_1.1.4       dplyr_1.0.7      
## 
## loaded via a namespace (and not attached):
##  [1] viridis_0.6.2         sass_0.4.0            jsonlite_1.7.2       
##  [4] viridisLite_0.4.0     foreach_1.5.1         bslib_0.3.1          
##  [7] assertthat_0.2.1      highr_0.9             stats4_4.1.0         
## [10] yaml_2.2.1            pillar_1.6.4          glue_1.5.0           
## [13] digest_0.6.28         RColorBrewer_1.1-2    colorspace_2.0-2     
## [16] htmltools_0.5.2       pkgconfig_2.0.3       GetoptLong_1.0.5     
## [19] purrr_0.3.4           scales_1.1.1          tibble_3.1.6         
## [22] farver_2.1.0          generics_0.1.1        IRanges_2.28.0       
## [25] ellipsis_0.3.2        withr_2.4.3           BiocGenerics_0.40.0  
## [28] cli_3.1.0             magrittr_2.0.1        crayon_1.4.2         
## [31] evaluate_0.14         fansi_0.5.0           doParallel_1.0.16    
## [34] tools_4.1.0           GlobalOptions_0.1.2   lifecycle_1.0.1      
## [37] matrixStats_0.61.0    ComplexHeatmap_2.10.0 stringr_1.4.0        
## [40] S4Vectors_0.32.2      munsell_0.5.0         cluster_2.1.2        
## [43] compiler_4.1.0        jquerylib_0.1.4       rlang_0.4.12         
## [46] iterators_1.0.13      rstudioapi_0.13       rjson_0.2.20         
## [49] circlize_0.4.13       labeling_0.4.2        rmarkdown_2.11       
## [52] gtable_0.3.0          codetools_0.2-18      DBI_1.1.2            
## [55] R6_2.5.1              gridExtra_2.3         knitr_1.36           
## [58] fastmap_1.1.0         utf8_1.2.2            clue_0.3-60          
## [61] dendextend_1.15.2     shape_1.4.6           stringi_1.7.5        
## [64] parallel_4.1.0        Rcpp_1.0.7            vctrs_0.3.8          
## [67] png_0.1-7             tidyselect_1.1.1      xfun_0.28