VT3D: a versatile visualization toolbox for 3D spatially resolved transcriptomic atlas
- Installation
- Quick Start
- Where can I download example datasets
- News: How can I browser my data in notbook with one line of python codes ?
- How can I use AtlasBroser to browse your data
- How can I create and visualize virtual slices
- How can I create MEP images
- How can I create surface models from omics data
- How can I adjust the coordinate of h5ad and mesh models by PCA
- How can I create continuous model by SVM
- Frequent Q & A
- Contact Us
- Cite Us
- Detailed Usages
A python3 environment with the following packages is required:
- json
- numpy
- pandas
- skimage
- sklearn
- anndata
- matplotlib
- seaborn
- pymeshfix [optional only for PVMesh]
- pyvista [optional only for PVMesh]
- meshio [optional only for PVMesh]
Note that we have not tested different versions of those packages yet, but below are the versions in our developing environment:
>>> json.__version__
'2.0.9'
>>> numpy.__version__
'1.24.2'
>>> pandas.__version__
'1.3.4'
>>> sklearn.__version__
'0.24.2'
>>> anndata.__version__
'0.8.0'
>>> matplotlib.__version__
'3.4.3'
>>> seaborn.__version__
'0.11.2'
git clone https://github.com/BGI-Qingdao/VT3D.git ./VT3D
All scripts are written in Python package. Thus, there is no need for any compilation.
This package contains compiled vt3d_browser. If your need the source codes, please go to https://github.com/BGI-Qingdao/VT3D_Browser
The basic input data is h5ad base and saves 3D coordinates in obsm. Surface models must be stored in .obj format.
All example datasets can be downloaded from http://www.bgiocean.com/vt3d_example/
In addition to the click-to-download mode from the above website, you can also download them by command line:
#--------- Developing drosophila embryos and larvae: E16-18h -----
mkdir E16_E18h
cd E16_E18h
wget -c http://www.bgiocean.com/vt3d_example/download/flysta3d/E16-18h_a_count_normal_stereoseq.h5ad
wget -c http://www.bgiocean.com/vt3d_example/download/flysta3d/E16-18h_a.tar.gz
wget -c http://www.bgiocean.com/vt3d_example/download/flysta3d/E16-18h.atlas.json
wget -c http://www.bgiocean.com/vt3d_example/download/flysta3d/fixed.json
tar -xzf E16-18h_a.tar.gz
cd ../
#--------- Developing drosophila embryos and larvae: L1
mkdir L1
cd L1
wget -c http://www.bgiocean.com/vt3d_example/download/flysta3d/L1_a_count_normal_stereoseq.h5ad
wget -c http://www.bgiocean.com/vt3d_example/download/flysta3d/L1_a.tar.gz
wget -c http://www.bgiocean.com/vt3d_example/download/flysta3d/L1_a.atlas.json
wget -c http://www.bgiocean.com/vt3d_example/download/flysta3d/fixed.json
tar -xzf L1_a.tar.gz
cd ../
#--------- Hypo Preoptic hypothalamic
mkdir hypo_preoptic
cd hypo_preoptic
wget -c http://www.bgiocean.com/vt3d_example/download/merfish/hypo_preoptic.h5ad
wget -c http://www.bgiocean.com/vt3d_example/download/merfish/mesh.tar.gz
wget -c http://www.bgiocean.com/vt3d_example/download/merfish/hypo_preoptic.atlas.json
wget -c http://www.bgiocean.com/vt3d_example/download/merfish/hypo_preoptic.coord.json
wget -c http://www.bgiocean.com/vt3d_example/download/merfish/hypo_preoptic.organ.json
tar -xzf mesh.tar.gz
cd ../
You can find all download commands from website
News: We are working on creating a python function to lauch atlas browser, please try this
You can find a step-by-step tutorial for building an atlas from website.
Here we use L1 atlas as an example:
cd L1
vt3d AtlasBrowser BuildAtlas -i L1_a_count_normal_stereoseq.h5ad -o L1_Atlas -c L1_a.atlas.json
cd L1_Atlas && vt3d AtlasBrowser LaunchAtlas -p 80
now open your web browser and try http://127.0.0.1/index.html
we use the E16-18h dataset as an example:
cd E16_E18h
vt3d AnySlice -i E16-18h_a_count_normal_stereoseq.h5ad -o test --spatial_key spatial --p0 '25,0,10', --p1 '50,30,0' --p2 '5,30,0' --thickness 1
cd E16_E18h
vt3d AnySlice -i E16-18h_a_count_normal_stereoseq.h5ad -o test_s5 --spatial_key spatial --p0 '25,0,10', --p1 '50,30,0' --p2 '5,30,0' --thickness 1 --slice_num 5
vt3d Auxiliary DrawSlices -i test_s5.h5ad -o drawan --color_by annotation
vt3d Auxiliary DrawSlices -i test_s5.h5ad -o drawrs --color_by slice_ID
vt3d Auxiliary DrawSlices -i test_s5.h5ad -o drawgn --color_by Acbp2
cd hypo_preoptic
vt3d MEP -i hypo_preoptic.h5ad --gene Baiap2 -o Baiap2
cd hypo_preoptic
vt3d MEP -i hypo_preoptic.h5ad -m Baiap2 -o Baiap2
Note: in pseudoFISH mode, you can use one of the following fluorescence colors
-m refers to magenta
-g refers to green
-y refers to yellow
-c refers to cyan
-r refers to red
-b refers to blue
First of all, the default view is APML plane (xy plane). You can switch to another view plane perpendicular to coordinate axes using --view=MLDV (yz plane) or --view=APDV (xz plane)
To get the 2D projection to an arbitrary plane defined by three non-collinear points, plesase use --view=[[0,0,0],[1,0,0],[1,1,0]]
cd hypo_preoptic
vt3d Auxiliary GrayScaleTIF -i hypo_preoptic.h5ad -o test3d -c hypo_preoptic.organ.json
Note: This action will generate hypo_preoptic.coord.json
The test3d.tif (opened by ImageJ 3DViewer) :
vt3d Auxiliary PVMesh -i E14-16h_a_count_normal_stereoseq.h5ad -o E14-16h_shell --spatial_key spatial --alpha 1.0
Here I use L1 larva data as input
>cat model.json
{
"Meshes" : {
"shell" : "L1_a/shell.obj",
"CNS" : "L1_a/CNS.obj",
"carcass" : "L1_a/carcass.obj",
"fatbody" : "L1_a/fatbody.obj",
"midgut" : "L1_a/midgut.obj",
"epidermis" : "L1_a/epidermis.obj",
"muscle" : "L1_a/muscle.obj",
"foregut" : "L1_a/foregut.obj",
"midgut_malpighian_tubule" : "L1_a/midgut_malpighian_tubule.obj",
"hindgut_malpighian_tubule" : "L1_a/hindgut_malpighian_tubule.obj"
},
"mesh_coord" : "fixed.json"
}
>vt3d Auxiliary PCA3D -i L1_a_count_normal_stereoseq.h5ad -o L1_pca --spatial_key spatial --model_json model.json
> cat conf.json
{
"spatial_key" : "spatial",
"annotation" : "annotation",
"step": 1,
"genes" : ["Acbp2",
"128up"
]
}
>vt3d Auxiliary BuildGrids -i E16-18h_a_count_normal_stereoseq.h5ad -o E16-18h_grids -m E16-18h_a_shell.obj -c conf.json
The default keyname for 3D spatial coordinates is spatial3D. Please make sure your coordinates are stored in obsm. You can use --spatial_key xxx to assign a different keyname.
Use fixed.json if you want to display the surface model in the same coordinate system of h5ad data. fixed.json
{"xmin": 0, "ymin": 0, "margin": 0, "zmin": 0, "binsize": 1}
Sometimes the surface model need to be shifted and scaled to fit the h5ad data, for example: hypo_preoptic.coord.json
{"xmin": -897, "ymin": -897, "xmax": 897, "ymax": 897, "margin": 10, "zmin": -290, "zmax": 260, "binsize": 50}
the surface model coordinate will be changed to
- multiple by 50
- x add -897
- y add -897
- z add -290
There are lots of ways for obtaining 3D coordinates of three non-collinear points. One straightforward way is to freely acquire coordinates using our 3D atlas browser.
-
First of all, your must assign more than two genes before Scoexp button can trigger.
-
Some genes may not exist in Scoexp matrix because thoes genes are too rare ( <3% cell ).
-
The Scoexp matrix are pre-computed by vt3d Auxiliary SCoexp. please manually move the output gene_scoexp folder into your atlas folder!
- Raising an issue is always a good way for questions
- Email us if you need: [email protected];[email protected];[email protected]
VT3D: a visualization toolbox for 3D transcriptomic data. Journal of Genetics and Genomics. 2023. https://doi.org/10.1016/j.jgg.2023.04.001
> vt3d -h
Usage : vt3d <action> [options ]
-h/--help show this short usage
Action:
-------------------------------------------------------------------------------
MEP Maximun Expression Projection. (project 3D structure to 2D)
AnySlice Extract a 2D slice from any given angle
AtlasBrowser Browse your data interactively
Auxiliary More auxiliary tools
-------------------------------------------------------------------------------
Detailed usage of each action:
vt3d <action> -h
Usage : vt3d MEP [options]
Options:
required options:
-i <input.h5ad>
-o <output prefix>
cmap mode
--gene geneid
[notice: enable --gene will override all pseudoFISH mode parameters]
--symbolsize [default 10, only used in cmap mode]
--cmap [default RdBu_r, only used in cmap mode]
pseudoFISH mode
-r [geneid that draw in Red(#ff0000) channel]
-g [geneid that draw in Green(#00ff00) channel]
-c [geneid that draw in Cyan(#00ffff) channel]
-m [geneid that draw in Magenta(#ff00ff) channel]
-y [geneid that draw in Yellow(#ffff00) channel]
-b [geneid that draw in Blue(#0000ff) channel]
optional configure options:
--binsize [default 5]
--spatial_key [defaut spatial3D, the keyname of coordinate array in obsm]
--view [default APML, must be APML/APDV/MLDV]
[APML -> xy plane]
[APDV -> xz plane]
[MLDV -> yz plane]
--plane [default '', example: "[[0,0,0],[1,0,0],[1,1,0]]" ]
[define any plane by three points]
[notice: enable --plane will override --view]
--drawborder [default 0, must be 1/0]
optional ROI options:
--xmin [default None]
--ymin [default None]
--zmin [default None]
--xmax [default None]
--ymax [default None]
--zmax [default None]
Example :
#example of cmap mode, will generate test.png
vt3d MEP -i in.h5ad -o test --gene wnt1 --view APML
#example of pseudoFISH mode, will generate test.tif :
vt3d MEP -i in.h5ad -o test -r notum -g wnt1 -m foxD --view APML
#example of cmap mode with assigned plane
vt3d MEP -i in.h5ad -o test --gene wnt1 --plane '[[0,0,0],[1,0,0],[1,1,0]]'
Usage : vt3d AnySlice [options]
Options:
required options:
-i <input.h5ad>
-o <output prefix>
--p0 <coordinate of p0>
--p1 <coordinate of p1>
--p2 <coordinate of p2>
optional options:
--thickness [default 16]
--slice_num [default 1]
--slice_step [default equal to thickness]
--spatial_key [default 'spatial3D', the keyname of coordinate array in obsm]
--X [default notset, create cross sections based on p1 only]
Example:
> vt3d AnySlice -i in.h5ad -o test --p0 "0,1,0" --p1 "1,0,0" --p2 "1,1,0"
> ls
test.h5ad
vt3d AtlasBrowser subaction:
BuildAtlas Generate cache files for WebCache action.
LaunchAtlas Start the atlas server for VT3D_Browse in the WebCache folder.
Usage: vt3d AtlasBrowser BuildAtlas [options]
Options:
required configurations:
-i <input.h5ad>
-c <conf.json>
-o [output prefix, default webcace]
Example:
> vt3d AtlasBrowser BuildAtlas -i in.h5ad -c atlas.json
> cat atlas.json
{
"Coordinate" : "spatial3D",
"Annotations" : [ "lineage" ],
"Meshes" : {
"body" : "example_data/body.obj" ,
"gut" : "example_data/gut.obj" ,
"neural" : "example_data/neural.obj" ,
"pharynx" : "example_data/pharynx.obj"
},
"mesh_coord" : "example_data/WT.coord.json",
"Genes" : [
"SMED30033583" ,
"SMED30011277" ,
... genes you want to display ...
"SMED30031463" ,
"SMED30033839"
]
}
Note:
Set "Genes" : ["all"] to export all genes in var.
The structure of the output atlas folder:
webcache
+---Anno
+---lineage.json
+---Gene
+---SMED30033583.json
+---SMED30011277.json
...
+---SMED30031463.json
+---SMED30033839.json
+---summary.json
+---gene.json
+---meshes.json
Usage: vt3d AtlasBrowser LaunchAtlas [options]
Options:
-p [port, default 80]
Example:
> vt3d WebServer
...
never stop until you press Ctrl-C
vt3d Auxiliary subaction:
GrayScaleTIF Generate 3D TIFF gray scale image as input of Slicer3D.
PVMesh Generate Mesh obj automaticly by pc.delaunay_3d
DrawVitrualSlices Draw 2D images groupby vitrual slice
PCA3D Reset 3d coordinate by PCA.
FormatMesh Shift and scale the mesh model.
BuildGrids Build grids( continuous model )
SCoexp Spatial-related co-expression matrix calculation.
Usage: vt3d Auxiliary GrayScaleTIF [options]
Options:
required configurations:
-i <input.h5ad>
-o <output prefix>
-c <conf.json>
--spatial_key [default 'spatial3D', the keyname of coordinate array in obsm]
Example:
> vt3d Auxiliary GrayScaleTIF -i in.h5ad -o test -c organ.json
> cat organ.json
{
"binsize" : 10,
"margin" : 10,
"keyname" : "annotation",
"targets": [
"all",
"Gut",
"Pharynx"
"Neural",
],
"grayvalue": [
50,
100,
150,
200
]
}
vt3d Auxiliary DrawVitrualSlices [options]
-i input h5ad
-o output prefix
--color_by gene name or annotation keywork
--vsid_key [default vsid, key of virtual slice id]
--coord2D [default spatial2D, keyname in obsm]
-t [default pdf, output type (png or pdf)]
-h/--help display this usage and exit
Usage : vt3d Auxiliary PVMesh [options]
Require:
pip install pymeshfix
pip install pyvista
pip install meshio
Options:
required options:
-i <input.h5ad>
-o <output prefix>
--target_cluster [default 'all']
--cluster_key [default 'clusters', the keyname of cell type in obs]
--spatial_key [default 'spatial3D', the keyname of coordinate in obsm]
--smooth 1000 [niter of Laplacian smoothing]
--alpha [default 2.0]
Notice: if target_cluster is all, cluster_key will be ignored.
Usage : vt3d Auxiliary PCA3D [options]
Options:
required options:
-i <input.h5ad>
-o <output prefix>
--spatial_key [default 'spatial3D', the keyname of coordinate in obsm]
--model_json [default None, the model.json]
Notice: data is priceless, no replace mode supported!
Example of model.json
{
"Meshes" : {
"body" : "example_data/body.obj" ,
"gut" : "example_data/gut.obj" ,
"nueral" : "example_data/neural.obj" ,
"pharynx" : "example_data/pharynx.obj"
},
"mesh_coord" : "example_data/WT.coord.json"
}
Usage : vt3d Auxiliary PCA3D [options]
Options:
required options:
-i <input.h5ad>
-o <output prefix>
-m <mesh.obj>
-c <conf.json>
Example of conf.json
{
"spatial_key" : "spatial",
"annotation" : "clusters",
"step": 10,
"genes" : [genename1,genename2]
}
Usage : vt3d Auxiliary SCoexp [options]
Options:
-i <input.h5ad>
-o <output prefix>
--spatial_key [default 'spatial3D', the keyname of coordinate in obsm]
--sigma [sigma of RBF kernel, default 15]
--genes [default None, file that contain target gene list. if None, all gene used]
Notice: too much genes will lead to huge memory cost and dist cost.