mapping scATAC-seq data with precomputated UMAP

[ccshao]

First thanks for the nice tool!

I was trying to mapping two scATAC-seq data, eg, ko to wt, with stream v1.0

In both datasets I used the precomputated UMAP projections. The codes for wt is shown bellow, in which the steps on add umap were taken from https://github.com/pinellolab/STREAM/issues/100#issuecomment-746414477.

import stream as st
import pandas as pd

st.__version__

st.set_figure_params(dpi=300,
                     style='white',
                     figsize=[5.4, 4.8],
                     rc={'image.cmap': 'viridis'})

adata = st.read(file_name='wt_counts.tsv.gz', workdir='./result_atac')
st.add_cell_labels(adata, file_name='wt_cell_label.tsv')
st.add_cell_colors(adata, file_name='wt_cell_label_col.tsv')

st.cal_qc(adata, assay='atac')
st.normalize(adata, method='tf_idf')
st.select_top_principal_components(adata, first_pc=False, n_pc=50, save_fig=True)

#- Precompuated umap.
umap_res = pd.read_csv("wt_umap.csv", sep="\t")

adata.obsm['X_dr'] = umap_res.to_numpy()
adata.obsm['X_vis_umap'] = umap_res.to_numpy()

st.plot_visualization_2D(adata,
                         n_neighbors=100,
                         color=['label'],
                         use_precomputed=True,
                         save_fig=True)

st.seed_elastic_principal_graph(adata, n_clusters=3, use_vis=True)
st.elastic_principal_graph(adata, epg_alpha=0.05)

However, when trying to align ko to wt:

adata_wt = st.read(file_name='CR_wt_atac_peaks.pkl')
adata_ko = st.read(file_name='CR_ko_atac_peaks.pkl')
adata_all = st.map_new_data(adata_wt, adata_ko)

Got the error:

Traceback (most recent call last): File "", line 1, in File "/home/shao/miniconda3/envs/env_stream/lib/python3.7/site-packages/stream/core.py", line 4040, in map_new_data feature = adata_ref.uns['params']['dimension_reduction']['feature'] KeyError: 'dimension_reduction'

In the source, I saw there is an transform step.

    if(method == 'umap'):
        if('trans_umap' in adata_ref.uns_keys()):
            trans = adata_ref.uns['trans_umap']
            adata_new.obsm['X_umap_mapping'] = trans.transform(input_data)
            adata_new.obsm['X_dr'] = adata_new.obsm['X_umap_mapping'].copy()
        else:
            raise Exception("Please run 'st.dimension_reduction()' using 'umap' first.")  

As I am using the precompuated umap, how could I instruct stream to use that projection?

Thanks and I am happy to provide more details if needed!