Abstract
Displaying data onto anatomical structures is a convenient technique to quickly observe tissue related information. However, drawing tissues is a complex task that requires both expertise in anatomy and the arts. While web based applications exist for displaying gene expression on anatograms, other non-genetic disciplines lack similar tools. Moreover, web based tools often lack the modularity associated with packages in programming languages, such as R. Here I present gganatogram, an R package used to plot modular species anatograms based on a combination of the graphical grammar of ggplot2 and the publicly available anatograms from the Expression Atlas. This combination allows for quick and easy, modular, and reproducible generation of anatograms. Using only one command and a data frame with tissue name, group, colour, and value, this tool enables the user to visualise specific human and mouse tissues with desired colours, grouped by a variable, or displaying a desired value, such as gene-expression, pharmacokinetics, or bacterial load across selected tissues. I hope that this tool will be useful by the wider community in biological sciences. Community members are welcome to submit additional anatograms, which can be incorporated into the package.
A stable version gganatogram has been deposited to neuroconductor, and a development version can be found on github/jespermaag/gganatogram.
Keywords: Anatograms, Anatomy, Tissues, Organs, ggplot2, R, Expression Atlas
Introduction
Efficiently displaying tissue information in multicellular organisms can be a laborious and time consuming process. Often researchers want to showcase differences in values, such as gene expression or pharmacokinetics between tissues in one organism, or between similar tissues in different groups.
Whereas bar charts and heatmaps provide an informative view of the differences between groups, it can be difficult to immediately observe the biological significance ( Figure 1a–b). As compared to an anatogram, where it is easy to quickly spot the differences between tissues or groups, and immediately provide biological context to these observations ( Figure 1c). This also has the added benefit that the audience, whether reading a paper or attending a lecture, will have to spend less time and effort to grasp the results.
Figure 1. Comparison between barplot (top left), heatmap (top right), and anatogram (bottom) to display tissue values between groups.
The values in the graphs are the same.
Several online tools to display gene expression in different tissues already exist 1– 4. Although these tools provide important information regarding gene expression in various tissues and organisms, other disciplines besides genetics are unable to utilise these applications due to the focus on genes. Moreover, these tools often only include a predefined set of experiments that can be visualised, leading to difficulties in presenting your own data. Other caveats with these tools are that it can be laborious to recreate the plot or automatically create plots from results.
Here I present gganatogram, an open source R package based on ggplot2 5 utilising the publicly available mouse and human anatograms from the Expression Atlas 1, 2. With this package it is easy for any R user to quickly visualise anatograms with specified colours, groups, and values. Using the familiar grammar from ggplot2 5, this program allows for modular anatograms to be generated.
Methods
Implementation
gganatogram is stored on neuroconductor 6, an open-source platform for rapid testing and dissemination of reproducible computational imaging software. A development version can be found on github/jespermaag/gganatogram, which allows for the community to post issues with the package, submit requests, or add anatograms by creating coordinate files.
source("https://neuroconductor.org/neurocLite.R") neuro_install("gganatogram", release = "stable", release_repo = latest_neuroc_release(release = "stable"))
The development version can be installed from github:
devtools::install_github("jespermaag/gganatogram")
Briefly, to generate the main list objects that contain all tissue coordinates, I downloaded SVG files from the Expression Atlas ( Available from gganatogram GitHub page 2) and processed them using a custom python script (available from GitHub). The script scraped through the SVG files to extract the name, coordinates, and SVG transformations. These were then post-processed in R to create the rda files that make up the tissue coordinates.
Operation
gganatogram requires an installation of R ≥3.0.0, ggplot2 5 v.3.0.0 and ggpolypath 7 v.0.1.0. The program should be able to run on any computer with the system requirements for R. Plots can be generated using a basic data.frame containing organ name, colour, type, or value, with the specified column names below. Organs are plotted one at a time based on the order of the data.frame. The tissue of each consecutive row will be layered on top of the previous. The gganatogram package provides four such data.frames containing all tissues available to plot, one for each human and mouse, and divided by sex.
hgMale_key, hgFemale_key, mmMale_key, mmFemale_key
These data frames have already specified colour, type, and an assigned random number to facilitate the start of plotting.
head(hgFemale_key) organ colour type value 1 pancreas orange digestion 10.373146 2 liver orange digestion 19.723172 3 colon orange digestion 14.853335 4 bone_marrow #41ab5d other 19.681587 5 urinary_bladder orange digestion 14.914273 6 stomach orange digestion 2.667599
The main function is called gganatogram(). By default, and without any arguments, it plots the outline of a male human with standard ggplot2 parameters. By adding just a few options, it is possible to quickly change to female, fill specified organs by selected colour, or fill the organs based on a value ( Figure 2).
Figure 2.
( A) Default plot generated by calling gganatogram(), ( B) adding female, plotting specified organs by ( C) colour, ( D) value.
library(gganatogram) library(gridExtra) organPlot <- data.frame(organ = c("heart", "leukocyte", "nerve", "brain", "liver", "stomach", "colon"), type = c("circulation", "circulation", "nervous␣system", "nervous␣system", "digestion", "digestion", "digestion"), colour = c("red", "red", "purple", "purple", "orange", "orange", "orange"), value = c(10, 5, 1, 8, 10, 5, 10), stringsAsFactors=F) A <- gganatogram() B <- gganatogram(fillOutline="#a6bddb", sex="female") + theme_void() C <- gganatogram(data=organPlot, fillOutline="#a6bddb", organism="human", sex="female", fill="colour")+ theme_void() D <- gganatogram(data=organPlot, fillOutline="#a6bddb", organism="human", sex="female", fill="value")+ theme_void() grid.arrange(A, B, C, D, ncol=4)
Use cases
This section provides additional plotting examples.
To plot all tissues per organism, use the provided key files that exist per organism and sex. This displays all tissues in the order of each data frame. To change the order in which organs are layered on top of each other, reorder the data frame to have those tissues at the bottom ( Figure 3).
Figure 3. Displaying all tissues available for human and mouse, male and female.
The colours are specified in the provided key data frames.
library(gganatogram) library(gridExtra) hgMale <- gganatogram(data=hgMale_key, fillOutline="#a6bddb", organism="human", sex="male", fill="colour") + theme_void() hgFemale <- gganatogram(data=hgFemale_key, fillOutline="#a6bddb", organism="human", sex="female", fill="colour") + theme_void()
mmMale <- gganatogram(data=mmMale_key, fillOutline="#a6bddb", organism="mouse", sex="male", fill="colour") + theme_void() mmFemale <- gganatogram(data=mmFemale_key, outline = T, fillOutline="#a6bddb", organism="mouse", sex="female", fill="colour") + theme_void() grid.arrange(hgMale, hgFemale, mmMale, mmFemale, ncol=4)
To compare anatograms, e.g. draw one specific anatogram side by side and compare values, a long table has to be created with the type column changed to the variables to compare. The following code recreates ( Figure 1c).
normal <- data.frame(organ = c("heart", "leukocyte", "nerve", "brain", "liver", "stomach", "colon"), value = c(10, 5, 1, 2, 2, 5, 5), type = rep("Normal", 7), stringsAsFactors=F) cancer <- data.frame(organ = c("heart", "leukocyte", "nerve", "brain", "liver", "stomach", "colon"), value = c(5, 5, 10, 12, 15, 5, 10), type = rep("Cancer", 7), stringsAsFactors=F) compareGroups <- rbind(normal, cancer) gganatogram(data=compareGroups, fillOutline="white", organism="human", sex="male", fill="value") + theme_void() + facet_wrap(~type) + scale_fill_gradient(low = "white", high = "steelblue")
Organs can also be separated by faceting, as per standard ggplot2 using facet_wrap ( Figure 4). This can help to display organs that are nested on top of each other.
Figure 4. Faceting tissues based on type and displaying the corresponding colour.
library(gganatogram) gganatogram(hgMale_key, fillOutline="#a6bddb", organism="human", sex="male", fill="colour") + theme_void() + facet_wrap(~type)
Because I elected to use ggplot2 5 for the package, the user can add additional layers from standard plots. This can be useful to show highlight features, such as metastasis, location of tissue biopsies, or gene expression of specific biopsies ( Figure 5).
Figure 5. Geom points added to a gganatogram to show the location of tissue biopsies (top left) along with a barplot of biopsy expression for an example gene (bottom).
Another option is to fill both tissues and points by value (top right). Red colour around plot added for emphasis.
library(gganatogram) library(dplyr) library(gridExtra) biopsies <- data.frame(biopsy = c("liver", "heart", "prostate", "stomach", "brain"), x = c(50, 55, 53, 60, 57), y = c(60, 48, 95, 68, 10), value = c(10, 15, 5, 2, 15)) p <- hgMale_key %>% dplyr::filter(organ %in% c("liver", "heart", "prostate", "stomach", "brain")) %>% gganatogram(fillOutline="lightgray", organism="human", sex="male", fill="colour") + theme_void() + ggtitle("Position of biopsies") p <- p + geom_point(data = biopsies, pch=21, size=2, aes(x =x, y = -y, fill = biopsy, colour= biopsy)) p2 <- ggplot(biopsies, aes(x = biopsy, y = value, fill = biopsy)) + geom_bar(stat= "identity", col="black") + theme_minimal() + theme(legend.position= "none")+ theme(axis.text.x = element_text(angle = 60, hjust = 1))+ ggtitle("Gene1␣expression") p3 <- hgMale_key%>% dplyr::filter(organ %in% c("liver", "heart", "prostate", "stomach", "brain"))%>% gganatogram(fillOutline="lightgray", organism="human", sex="male", fill="value") + theme_void() + ggtitle("Value␣of␣biopsies") + geom_point(data = biopsies, pch=21, size=3, aes(x =x, y = -y, fill = value), colour="red") lay <- rbind(c(1,2), c(1,2), c(3, NULL)) grid.arrange(p, p3, p2, layout_matrix = lay)
Summary
In summary, I have designed and implemented an R package to easily visualise anatograms based on ggplot2 5 and the anatograms from Expression Atlas 2, which when combined create a powerful tool to plot and display tissue information.
The one line command to generate these plots should allow for users with even limited R knowledge to create informative anatograms for publications or presentations.
Data availability
SVG files are available from GitHub: https://github.com/ebi-gene-expression-group/anatomogram/tree/master/src/svg
Software availability
-
1.
Link to version control repository containing the source code:
-
2.
Link to development version:
-
3.
Link to archived source code as at time of publication:
http://doi.org/10.5281/zenodo.1434233 8
Software license: GPL-2
Acknowledgments
I would like to thank the neuroconductor team: Ciprian Crainiceanu, John Muschelli, Brian Caffo, and Adi Gherman for storing gganatogram on their repository.
Paul Brennan ( @brennanpcardiff) for adding additional checks to the package.
I would like to thank Irene Papatheodorou and the Expression Atlas team at EMBL-EBI for making the anatograms available.
I would also like to thank Anna Antoniak for editing the manuscript, and Stephen Rudley for manuscript feedback.
Funding Statement
The author(s) declared that no grants were involved in supporting this work.
[version 1; referees: 2 approved]
References
- 1. Papatheodorou I, Fonseca NA, Keays M, et al. : Expression Atlas: gene and protein expression across multiple studies and organisms. Nucleic Acids Res. 2018;46(D1):D246–D251. 10.1093/nar/gkx1158 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Petryszak R, Keays M, Tang YA, et al. : Expression Atlas update--an integrated database of gene and protein expression in humans, animals and plants. Nucleic Acids Res. 2016;44(D1):D746–D752. 10.1093/nar/gkv1045 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Lekschas F, Stachelscheid H, Seltmann S, et al. : Semantic Body Browser: graphical exploration of an organism and spatially resolved expression data visualization. Bioinformatics. 2015;31(5):794–796. 10.1093/bioinformatics/btu707 [DOI] [PubMed] [Google Scholar]
- 4. Palasca O, Santos A, Stolte C, et al. : TISSUES 2.0: an integrative web resource on mammalian tissue expression. Database (Oxford). 2018;2018. 10.1093/database/bay003 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Wickham H: ggplot2: Elegant Graphics for Data Analysis. Springer-Verlag New York,2016; ISBN 978-3-319-24277-4. 10.1007/978-3-319-24277-4 [DOI] [Google Scholar]
- 6. Muschelli J, Gherman A, Fortin JP, et al. : Neuroconductor: an R platform for medical imaging analysis. Biostatistics. 2018. 10.1093/biostatistics/kxx068 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Sumner MD: ggpolypath: Polygons with Holes for the Grammar of Graphics. Reference Source [Google Scholar]
- 8. Maag J, Muschelli J, Brennan P, et al. : jespermaag/gganatogram: First release (Version V1.0.0). Zenodo. 2018. 10.5281/zenodo.1434233 [DOI] [Google Scholar]