In AllenInstitute/atlasplot: Easy Plotting of the Allen Brain Altases

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

Overview

This package aims to simplify the process of producing graphics for the expression data available at the Allen Brain Map. Specifically, it provides functions for plotting gene expression from six atlases:

1. Allen Human Brain Atlas - atlasplot generates bar plots of gene expression for each brain structure, as shown in Hawrylycz, et al 2012 and Hawrylycz, et al 2015. Data is stored in the companion R package hbadata.
2. BrainSpan: Prenatal LMD microarray - atlasplot generates several plots showing gene expression in prenatal (15-21 PCW) human brain, as shown in Miller et al 2014. Data is stored in fetdata.
3. BrainSpan: Developmental Transcriptome - atlasplot generates bar plots of gene expression for each brain structure across donors (prenatal through young adult). Data is stored in brspdata.
4. NIH Blueprint NHP Atlas -atlasplot generates several plots showing gene expression in non-human primate brain (spanning multiple brain regions and pre- and post-natal time points), as shown in Bakken et al 2016. Data is stored in nhpdata.
5. Allen Mouse Brain Atlas - atlasplot generates bar plots of gene expression in the adult mouse brain by querying the Allen Institute API.
6. Allen Developing Mouse Brain Atlas - atlasplot generates bar plots of gene expression in the developing mouse brain by querying the Allen Institute API.

Setup

First, install atlasplot and the associated data packages, if needed.

# Install devtools package
install.package("devtools")

# Install atlasplot (with this vignette)
devtools::install_github("AllenInstitute/atlasplot", build_vignettes = TRUE)

# Install data packages (note, these are big files!)
devtools::install_github("AllenInstitute/hbadata")
devtools::install_github("AllenInstitute/fetdata")
devtools::install_github("AllenInstitute/brspdata")
devtools::install_github("AllenInstitute/nhpdata")

Next, load atlasplot library. The dependencies should be loaded automatically.

library(atlasplot)

Finally, change your working directory, if needed. By default most functions save plots as pdf files in your working directory (which ensures they retain the optimal height and width), but can be directed to the standard output if desired.

Generate some plots

The remainder of this vignette shows you how to generate plots for each of the six atlases described above, using some intersting genes as examples. This R library uses caching, which means that functions may run extremely slowly the first time you call them, but should run relatively more quickly in subsequent function calls.

Allen Human Brain Atlas

The Allen Human Brain Atlas includes a study that measures gene expression on several hundred brain locations across 6 donors using bulk and microdissected tissue. The function hba_subregions_plot generates bar plots of gene expression for each brain structure in adult human brain. This can be used to visualize gene expression across the brain. For this example, let's look at PROX1, which is one of the best markers for hippocampus dentate gyrus (DG) granule cells in mouse.

hba_subregions_plot("PROX1")

This code should put a file called "PROX1_HBA_subregionsPlot.pdf" in your working directory. In this file each plot represents gene expression in a single brain (6 total), and each bar within a plot represents the set of microarray samples collected from the brain region listed below the bar. Brain regions correspond to the abbreviations from the Human Brain Reference Atlas. As expected, PROX1 has higher expression in DG than in other hippocampus brain regions (adjacent red bars), along with some thalamic structures (orange bars a bit to the right).

BrainSpan Atlas of the Developing Human Brain

This library generates plots for two separate studies within the BrainSpan project. The first is the Prenatal LMD microarray, which is a study similar in design to (and using the same microarray as) the Allen Human Brain Atlas. It looks at a couple hundred brain structures across four brains in prenatal (15-21 postconceptual weeks, or PCW) human. The second is the Developmental Transcriptome study, which uses RNA-seq and exon arrays to study 16 brain regions across prenatal and postnatal development (through young adulthood).

Prenatal LMD microarray

In this study, let's focus on FOXP2, which is a transcription factor implicated in the specialization of language areas, and which is found in deep layer neurons in adult cortex. Using fet_brain_location_expression_plot and fet_cortex_expression2D_plot, we can reproduce a version of the figure shown in Extended Data Figure 10 from Miller et al 2014.

fet_cortex_expression2D_plot("FOXP2")
fet_brain_location_expression_plot("FOXP2")

"FOXP2_fetalHuman_2DPlotInNeocortex.pdf" shows a two dimensional heatmap of expression, one per brain, where the rows corresponds to prenatal cortical layers and the columns correspond to brain structures. From this we can see that FOXP2 shows differences across layer and region, with high expression in progenitor cell layers (SZ/VZ) in frontal cortex, and in deep cortical layers (CPi/SP) in the other three cortical lobes. The second plot ("FOXP2_fetalHuman_BrainLocationExpressionPlotsForEachLayer.pdf") shows the same information in a different way, with each row of plots corresponding to cortical layers and each column corresponding to different brains. Within a plot the position of the dots correspond to the centroid of the brain region shown in the context of an image of the brain, and the color and size of the dots correspond to expression in that region.

There is also a function (fet_subregions_plot) which can plot barplots in the same format as for the adult human. Let's see whether PROX1 is expressed in DG and thalamus in prenatal cortex.

fet_subregions_plot("PROX1")

As shown above in adult, we can see in "PROX1_fetalHuman_structureBarPlotAll.pdf" that PROX1 has a similar expression pattern already by 15 PCW.

Developmental Transcriptome

The function brsp_subregions_plot generates bar plots of gene expression for each brain structure across donors (prenatal through young adult), outputting one plot sorted first by brain region and one sorted first by age. In this example, let's plot SLC17A7, which is a canonical marker for cortical neurons to see what brain region it is expressed in and at which ages the gene is expressed.

# The default is to show exon array data
brsp_subregions_plot("SLC17A7")
# We can also plot the RNA-seq data 
brsp_subregions_plot("SLC17A7","rna")

We can see from the top plot that SLC17A7 is expressed in cortex, cerebellum and amygdala but not striatum or thalamus. In addition, expression levels of this gene increase from low to high levels around late prenatal or perinatal development.

NIH Blueprint Non-Human Primate Atlas

The NIH Blueprint NHP Atlas contains a map gene expression it rhesus monkey at high spatial and temporal resolution. Specifically, microarrays were used to measure gene expression on laser microdissected sections from every available cortical layer in two cortical areas (along with other select structures in neocortex, hippocampus, striatum, amygdala, and thalamus) across six prenatal and four postnatal time points.

atlasplot provides three functions for displaying gene expression in neocortex. For illustrative purposes, let's focus on two genes presented in Bakken et al 2015: EOMES (also called TBR2), which is a marker for intermediate progenitors, and MOBP, which is a marker for a mature oligodendrocytes.

## Plots for EOMES
nhp_cortex_expression2D_small_plot("EOMES")
nhp_cortex_expression2D_plot("EOMES")
nhp_cortex_series_plot("EOMES")

## Plots for MOBP
nhp_cortex_expression2D_small_plot("MOBP")
nhp_cortex_expression2D_plot("MOBP")
nhp_cortex_series_plot("MOBP")

Both expression2D plots show gene expression summaries with rows corresponding to cortical layer and columns corresponding to age. The difference is the small plot shows expression averaged across all cortical structures, while the main function shows individual points for each region. From these plots we can see that EOMES is most highly expressed at E90 and before in progenitor layers (particularly SZ), while MOBP is expressed postnatally across all layers with highest expression in the white matter (WM) and adjacent L6. Finally, nhp_cortex_series_plot outputs line plots focused on comparing gene expression across cortical areas. Each box is a layer, with lines corresponding to brain regions and points corresponding to the time series. For example, we can see that at in L5 and L6 of V1, MOBP expression levels have reached their peak in E0, while in ACG they don't reach their peak until E3. In contrast, expression levels of EOMES start declining in ACG at earlier prenatal stages than in V1. These results recapitulate prior observations on differing rates of neurogenesis and circuit maturation in different cortical areas.

Allen Mouse Brain Atlases

While the primary focus of atlasplot is to present gene expression plots of the Allen Brain Map microarray and RNA-seq data sets that represent the data in more biologically relevant contexts than the default 1-dimensional heatmap used at brain-map.org, we also provide some functions for plotting average expression energy in the developing and adult mouse. These plots pull data directly from the Allen Brain Atlas API and then plot the results using bar plots of the formats described above. Recall that if this is your first time using this function, the function calls may be quite slow until cache is established.

Adult mouse

The function mouse_subregions_plot makes these plots. For adult, set atlas="Mouse". This function has two additional important parameters. The experiment parameter will allow you to directly specify which Experiment ID to pull data from, which is particularly relevant for genes with more than one experiment (e.g., genes with both saggital and coronal sections). If this is left blank you'll be given choices to type in later. The struct_depth parameter let's you pull at a specific level of the ontology. Higher numbers mean smaller, more specific brain regions while lower numbers are larger brain areas. In it often useful to make plots at multiple depths since the ontology has different depths in different parts of the brain.

For this example, let's revisit Prox1 to see whether the expression in mouse matches the expression in human (dentate gyrus + thalamus).

mouse_subregions_plot(
  "Prox1",
  atlas = "Mouse",  
  struct_depth = 3
)

Here I chose the coronal section (73520980), but the structure depth was too low, as seen in the file Prox1_Mouse_3_brainRegionBarplot_ExpressionEnergies.pdf in your working directory. Let's try again at higher depth, and with the experiment specified in advance. We also need to adjust the image width so all the labels show.

mouse_subregions_plot(
  "Prox1",
  atlas = "Mouse",  # Or "DevMouse"
  experiment = 73520980,
  struct_depth = 7,
  im_width = 60
)

As expected, the highest bar corresponds to dentate gyrus.

Developing mouse

The same function is used to pull gene expression in developing mouse, with the expection that the experiment ID does not need to be specified (all experiments for a given gene are used), and we need to set atlas = "DevMouse". In this example, let's see how Eomes looks across mouse development.

mouse_subregions_plot(
  "Eomes",
  atlas = "DevMouse",
  struct_depth = 5
)

The anatolical terms are hard to interpret, but as in NHP, the highest expression of this gene is found prenatally rather than postnatally.

Developmental time series comparison

The final plotting function is species_expression_series_plot, which plots neocortical gene expression values for a time series of ages for human, non-human primate, mouse, and rat for the subset of genes in which data for at least two species is available. Use UPPER CASE gene symbols (as in human) for this function. For this example, let's take two canonical marker genes for GABAergic interneurons: PVALB and VIP

species_expression_series_plot("PVALB")
species_expression_series_plot("VIP")

The output of this plot is a series of points and associated best fit curves for up to four of the species. The x-axis is escore, which is a metric of age that aligns significant developmental milestones across species, following the algorithm for translating time developed by Barbara Finlay and colleagues. The y-axis is a normalized metric for gene expression. Thus a plot where two lines overlap would indicate that a gene shows comparable temporal patterning across species. In this case, VIP shows a relatively comparable rate of progression from low to high expression across development across species, whereas PVALB shows a slower, later increase in primate than in rodent species, with expression in human still increasing into postnatal development (escore > 1).

sessionInfo()

AllenInstitute/atlasplot documentation built on Aug. 11, 2021, 1:52 a.m.