Usage

Load data form h5ad file

If you have 2 anndata object that already finished scaling, normalizing,… You can just simply use screg2.RegNMF() to integrate rna and atac data, with batch remove. screg2.RegNMF() will return the rna anndata object with scReg_reduction in it rna_adata.obsm

import pandas as pd
import numpy as np
import scanpy as sc
import screg2

rna_file = '/path/to/your/rna/h5ad/file'
atac_file = '/path/to/your/atac/h5ad/file'
rna_adata = sc.read_h5ad(rna_file)
atac_adata= sc.read_h5ad(atac_file)
rna_adata = screg2.RegNMF(rna_data=rna_adata, atac_data=atac_adata, Meta_data=rna_adata.obs,batch_type='sample', maxiter=100, key_added="scReg_reduction")

screg2.RegNMF() will return the rna anndata object with scReg_reduction in it rna_adata.obsm

rna_adata.obsm['scReg_reduction']

array([[4.13353097e-14, 1.08500913e-10, 3.27313772e-11, ...,
        0.00000000e+00, 0.00000000e+00, 0.00000000e+00],
      [2.09097071e-14, 1.96949657e-16, 1.67404740e-11, ...,
        0.00000000e+00, 0.00000000e+00, 0.00000000e+00],
      [1.12355639e-10, 3.41268614e-11, 6.89114276e-11, ...,
        0.00000000e+00, 0.00000000e+00, 0.00000000e+00],
      ...,
      [6.38938638e-14, 7.88140180e-11, 5.65415185e-13, ...,
        0.00000000e+00, 0.00000000e+00, 0.00000000e+00],
      [1.88311115e-10, 1.70259133e-10, 6.51093299e-12, ...,
        0.00000000e+00, 0.00000000e+00, 0.00000000e+00],
      [2.15576528e-12, 9.40752216e-11, 1.57056569e-12, ...,
        0.00000000e+00, 0.00000000e+00, 0.00000000e+00]])

Warning

If you do not have the meta data of cells for their sample infomation the nex code might help you

rna_adata.obs['sample'] = adata.obs_names.str.split("-").str[1]

Clustering and Umap

sc.pp.neighbors(rna_adata, use_rep='scReg_reduction', key_added='scReg_neighbors', n_neighbors=40)
sc.tl.leiden(rna_adata, neighbors_key='scReg_neighbors', key_added='scReg_leiden')
sc.tl.umap(rna_adata, neighbors_key='scReg_neighbors')
sc.pl.umap(rna_adata, color="scReg_leiden", legend_loc="on data",legend_fontsize="small",size=6, title="scReg")
# If you want to save the figure you can
# sc.pl.umap(rna_adata, color="scReg_leiden", legend_loc="on data",legend_fontsize="small",size=6, save="_scReg.pdf" title="scReg")

Load data from h5 file

import scanpy as sc
h5_file = '/path/to/your/h5/file'
adata = sc.read_10x_h5(h5_file,gex_only=False)
# Make sample name Meta data. Makesure your cell barcode are end with -sample_name like this "AATTAATT-34"
adata.obs['sample'] = adata.obs_names.str.split("-").str[1]

Create RNA and ATAC Anndata object

#start filtering cell by rna rna_adata= adata[:,adata.var[‘feature_types’]==’Gene Expression’] atac_adata = adata[:,toyData.var[‘feature_types’]==’Peaks’] atac_adata =

import pandas as pd
import numpy as np
import scanpy as sc
import screg2
file = "/path/to/your/h5/file"
adata = sc.read_10x_mtx(path='/data2/duren_lab/palmetto/cham/toyData/filtered_feature_bc_matrix/',cache=True,gex_only=False)
rna_adata = adata[:,toyData.var['feature_types']=='Gene Expression']
atac_adata = adata[:,toyData.var['feature_types']=='Peaks']
sc.pp.normalize_total(rna_adata)
sc.pp.normalize_total(atac_adata)



# This part will return redaction to adata_rna
adata_rna = screg2.RegNMF(E = rna_adata, O =atac_adata, batch_type = "sample")