Skip to main content

No project description provided

Project description

https://badge.fury.io/py/svgbit.svg

For further detail, please visit our API reference on readthedocs.org

Installation

Install svgbit with pip:

pip install svgbit

Command-line Interface

svgbit has a command line version. Just tape:

svgbit --help

after installation, and you may get a short help massage:

usage: svgbit [-h] [--k K] [--n_svgs N_SVGS] [--n_svg_clusters N_SVG_CLUSTERS]
              [--he_image HE_IMAGE] [--save_dir SAVE_DIR] [--cores CORES] read_path

Find spatial variable genes for Spatial Trasncriptomics data.

positional arguments:
read_path             Read Spatial Transcriptomics data. Support format in 10X
                      Space Ranger(dir named ``outs``) and anndata hdf5


optional arguments:
-h, --help            show this help message and exit
--normalization NORMALIZATION
                      apply which normalization on read data. If None (default),
                      neither normalization will apply. Supported values: None, cpm,
                      logcpm (default: None)

--k K                 number of nearest neighbors for KNN network (default: 6)
--n_svgs N_SVGS       number of SVGs to find clusters (default: 1000)
--n_svg_clusters N_SVG_CLUSTERS
number of SVG clusters to find (default: 8)
--he_image HE_IMAGE   path to H&E image. Only used for visualization (default: None)
--save_dir SAVE_DIR   path to save results (default: .)
--cores CORES         number of threads to run svgbit (default: 8)

Follow the introduction and results will save to –savedir.

Python Interface

svgbit has a set of python API. You may run svgbit through command line or python. We recommend the usage of python API for more feature and convient control of your input data.

Load Data

svgbit could load data from Space Ranger output directory:

import svgbit as sb
dataset = sb.load_10X("spaceranger_output/outs")

Or load data from csv files:

import svgbit as sb
dataset = sb.STDataset(
    count_df="Data/count_df.csv",
    coordinate_df="Data/coor_df.csv",
    count_df_kwargs={"index_col": 0, "header": 0},
    coordinate_df_kwargs={"index_col": 0, "header": 0},
)

Data Preprocessing

Genes with low variance may filter with:

dataset = sb.filters.low_variance_filter(dataset, var=0)

Genes with extremely high expressions usually show no pattern and may distrub performing. Filter with:

dataset = sb.filters.quantile_filter(dataset, 0.99)

svgbit alse has count normalization functions:

dataset = sb.normalizers.logcpm_normailzer(dataset)

Feel free for choosing gene filters and data normalizers. Other filters and normalizers are also provided. Visit our API reference for further detail.

Run svgbit

To perform full pipeline of svgbit, running:

sb.run(dataset)

Visualization

Draw SVG heatmap and spot type distribution map with:

sb.svg_heatmap(dataset, save_path="heatmap.jpg", he_image="he_image.jpg")
sb.spot_type_map(dataset, save_path="spot_type.jpg", he_image="he_image.jpg")

Parameter he_image is optional. If not specified, hotspot discription map will show without morphological information.

Details about sb.run()

When you perform sb.run(), sevaral steps will be done as below. For further detail of calculation, please refer to our publication.

Acquire weight

To calculate hotspot matrix, svgbit needs a weight network which discribes association across spots. svgbit uses k-nearest neighbors with 6 neighbors as a default. You may pass key word argument k to sb.run() to change this behavior.

In this step, sb.run() will execute STDataset.acquire_weight() method with given parameters. You may also perform this step by:

dataset.acquire_weight()

Weight will save as attribute weight of STDataset and detailed discription of weight is saved to weight_type attribute. Users may provide a libpysal.weights.W instance as user-specified weight:

dataset.weight = user_specified_weight

Acquire hotspot

Hotspot matrix is estimated by:

dataset.acquire_hotspot()

and save to hotspot_df attribute.

Density

AI and Di value discribed in our paper will be calculate by:

dataset.acquire_density()

and save to AI and Di attribute as pd.Series.

Find SVG clusters

SVG clusters is estimated by:

dataset.find_clusters()

and save to svg_cluster attribute.

For further discription of hotspot, AI, Di and SVG cluster, please refer to our manuscript.

Citation

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

svgbit-0.2.10.tar.gz (31.8 kB view hashes)

Uploaded Source

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page