pairpot 0.0.3

Pairpot: a database with real-time lasso-based analysis tailored for paired single-cell and spatial transcriptomics.

Usage of Pairpot Library

pairpot library provides some powerful tools to analyze single-cell data and spatial transcriptomics data.

pairpotlpa library provides support of lassoView function of pairpot.

Run in console to install pairpot and pairpotlpa.

pip install pairpot

And use this code to import.

import pairpot as pt

You can also find our library code in https://github.com/lyotvincent/Pairpot/tree/master/backend/pairpot.

Download Dataset from Pairpot

Corresponding adata can be downloaded in MetaInfo or use the code as following.

pt.download acceptes 3 parameters: dataset_id, type and file.dataset_id acceptes a string denoting the id of dataset in pairpot, type denotes the data type ('sc': single-cell dataand 'sp': spatial transcriptomics data) and file denotes which file('complete': the complete file and 'meta': the processed file) needs to be download.

pt.download returns adata with anndata format. Following code shows the usage of pt.download

# Use complete dataset_id to download
adata = pt.download("STDS0000235", type='sc', file='complete')

# Also, you can use the last 3 digits instead.
adata = pt.download("235", type='sc', file='complete')

Perform Lasso-View Using Python Offline

pt.lassoView can perform Lasso-View for anndata offline.

pt.lassoView acceptes parameters adata, selected, formatand do_correct.adata denotes the anndata format data or the path of h5ad file to be analyzed. selected acceptes a list to denote the lassoed cells indexs, which can also be gotten by lasso tools on the pairpot website. format denotes the format of adata, 'h5adfile' denotes that parameter adata is a path of h5ad file and 'anndata' denotes that parameter adata is an anndata format data. do_correct denotes that after the label-propagation, does Lasso-View conduct knn rectification or not. In default, format is set to 'anndata' and do_correct is set to True. pt.lassoView return a list of cell indexs, including ones are considered as the same cell type of the lassoed parts.

Following code shows the usage of pt.lassoView.

# Perform Pair-View using Python
# Run in console: pip install pairpot
import pairpot as pt

# Corresponding adata can be downloaded in MetaInfo or use the code as following.
adata = ad.read_h5ad("sample.h5ad")

lassoed_index = # Which can be generated in the pairpot website.
[70,153,166,290,343,423,513,537,
863,957,963,1111,1207,1252,1308,
1337,1354,1547,1750,1908,1928,1962,
1984,2043,2178,2211,2305,2343,2346,
2350,2427,2866,2932,2969]

refined_index = pt.lassoView(lassoed_index, adata)

Perform Pair-View Using Python Offline

pt.pairView can perform PairView for anndata offline.

Similar to pt.lassoView, pt.pairView acceptes parameters selected, scdata, spdata, format.

selected acceptes a list to denote the selected cells indexs, which can also be gotten on the pairpot website. scdata and spdata acceptes the anndata format single-cell and spatial transcriptomic data or the path of sc and sp h5ad file. format controls the format of scdata, and spdata: 'h5adfile' denotes that parameter scdata and spdata are both path of h5ad files and 'anndata' denotes that parameter scdata and spdata are both anndata format datas.

In default, format is set to 'anndata'. pt.pairView return a list of proportion o f selected cells in each spot.

Following code shows the usage of pt.pairView.

# Perform Pair-View using Python
# Run in console: pip install pairpot
import pairpot as pt

# Corresponding adata can be downloaded in MetaInfo or use the code as following.
scdata = pt.download("STDS0000235", type='sc', file='complete')
spdata = pt.download("STDS0000235", type='sp', file='complete')
    
# Code for generating cell proportions for "your annotation" in SRT data offline
lassoed_index = 
[993,1991,2136,2685,2731,70,112,
153,166,290,343,423,513,537,863,
957,963,1111,1132,1207,1252,1308,
1337,1354,1547,1750,1908,1928,
1962,1984,2043,2131,2178,2211,
2292,2305,2343,2346,2350,2382,
2427,2486,2715,2866,2932,2969,2992]
props = pt.pairView(lassoed_index, scdata, spdata)

# the cell proportions inferred online can be found through 'Save' -> 'JSON'.
spdata.obs["your_annotation"] = props

Using Online Refined Results Directly

As follow, pairpot website also provides refined_index in Lasso-View tool. which can be used directly to conducted further research.

refined_index = 
[42,46,70,112,153,166,290,343,405,
423,513,518,537,710,863,942,957,
963,1085,1111,1132,1207,1252,1308,
1337,1354,1547,1703,1712,1750,1908,
1928,1962,1984,2043,2065,2131,2178,
2211,2292,2305,2343,2346,2350,2382,
2427,2715,2747,2866,2932,2944,2969,2992]

adata.obs['annotation'] = list(adata.obs['annotation'])
adata.obs['annotation'].iloc[refined_index] = "annotation 1"
adata.obs['annotation'] = adata.obs['annotation'].astype("category")

Offline Data Processing

Pairpot library also provides some functions to curate data before conducting analysis.

Preprocessing, Data Integration and Neighbor Graph Construction

pt.lassoProc function conducts data preprocess as follows:

• Removing minority of cells
• Performing PCA
• Establishes similarity metrix $W$ and probability transition matrix $P$
• Perform dimensional reduction for visulization (using UMAP and t-SNE based on the neighbor graphs)
• Perform multi-scale celluar context representations of SRT data by MENDER

pt.lassoProc accepts only one parameter path, denotes the directory path. Then lassoProc process all h5 or h5ad file in the directory and save the processed file as New_xxx.h5ad.

Following code shows the usage of pt.pairView.

pt.lassoProc('./download')

UCell Scores Evaluation

pt.pairProc function conducts offline clustering as follows:

• Segmenting the spatial domains in multiple slices using MENDER based on the multi-scale celluar context representations
• For scRNA-seq data, clusting cells using Leiden algorithm with the default resolution
• Using UCell to assign signature scores and evaluate cell types for datasets lacking original annotations

pt.pairProc accepts parameters: adata, organs, top, alpha, n_jobs and clu_key. adata denotes an anndata format data to be processed, organs denotes the organs and species from existing cell marker databases(like PanglaoDB and CellMarker ), top denotes the topK ratio, alpha denotes the resolution level, n_jobs denotes the number of threads for parallel processing and clu_key denotes the cluster approach.

In default, top, alpha and n_jobs are separatedly set to 0.05, 10e-40 and 16, and clu_key are set to 'leiden-1'.

Following code shows the usage of pt.pairProc.

adata=ad.read_h5ad('sample.h5ad')
organs=['Brain','Blood']
proc_adata=pt.pairProc(adata,organs)