Question about data access for Parse Biosciences Sketch Integration tutorial

[mbm2224]

Hello, I have a question about how the data were accessed in Sketch integration using a 1 million cell dataset from Parse Biosciences vignette. The tutorial states that the dataset from Parse Biosciences is downloaded as an h5.ad file.

I followed the link to Using BPCells with Seurat Objects vignette to see how the data were downloaded and processed. The download link for the data takes me to a Parse Biosciences webpage. There are instructions to download the genes, matrix, and metadata file, but no option for an h5.ad file.

How was this file: 'ParseBio_PBMC.h5ad' obtained for this line of code

parse.data <- open_matrix_anndata_hdf5(
  "/brahms/hartmana/vignette_data/h5ad_files/ParseBio_PBMC.h5ad"

from Using BPCells with Seurat Objects vignette? Did you follow the steps for preprocessing from Parse Biosciences? If so, how far in the tutorial did you go until you wrote out the h5.ad file?

Thanks in advance for clarifying!

[mdidish]

Hello, have youf foud the answer?

I tried this:
1st with Python:

import scanpy as sc
adata = sc.read_mtx('DGE_1M_PBMC.mtx')
adata.write('ParseBio_PBMC.h5ad')

2nd with R

parse.data <- open_matrix_10x_hdf5(
  path = "../RawData/ParseBio_PBMC.h5ad"
)

But I obtained this error:
Loading multi-genome matrices from old-style 10x hdf5 files is unsupported

[ashwini-girish]

In case you are still looking for the dataset, you can download it using this link. The same is now linked in the article as well.

[mbm2224]

Thank you!

[Pancreas-Pratik]

For reference, I was having some trouble for re-creating the initial dataset used in the Parse Biosciences Sketch Integration tutorial, but this is how I was able to do it:

From the Seurat vignette section OP mentioned (https://satijalab.org/seurat/articles/seurat5_bpcells_interaction_vignette.html#parse-biosciences), I clicked the here where in that section it says: "The data is available for download here.

I then did this part of the python code:

import os
import sys
import scipy
import numpy as np
import pandas as pd
import scanpy as sc
import seaborn as sns
import scipy.io as sio
import scanpy.external as sce
import matplotlib.pyplot as plt

sc.settings.verbosity = 1
sc.settings.figdir = '/my-figure-directory/'
obj_save_path = '/my-object-save-path/'
mat_path = '/my-expression-matrix-path/'

# Adjust Scanpy figure defaults
sc.settings.set_figure_params(dpi=100, fontsize=10, dpi_save=400,
    facecolor = 'white', figsize=(6,6), format='png')

and also this part, but instead of writing the adata_obj1.h5ad immediately after the sc.read_mtx() line. I moved it after the adding cell meta data. So basically the resulting .h5ad now also has cell and gene names.

# The DGE_filtered folder contains the expression matrix, genes, and files
adata = sc.read_mtx(mat_path + 'DGE_1M_PBMC.mtx')


# adata = sc.read(obj_save_path + 'adata_obj1.h5ad')

# reading in gene and cell data
gene_data = pd.read_csv(mat_path + 'all_genes_1M_PBMC.csv')
cell_meta = pd.read_csv(mat_path + 'cell_metadata_1M_PBMC.csv')

# find genes with nan values and filter
gene_data = gene_data[gene_data.gene_name.notnull()]
notNa = gene_data.index
notNa = notNa.to_list()

# remove genes with nan values and assign gene names
adata = adata[:,notNa]
adata.var = gene_data
adata.var.set_index('gene_name', inplace=True)
adata.var.index.name = None
adata.var_names_make_unique()

# add cell meta data to anndata object
adata.obs = cell_meta
adata.obs.set_index('bc_wells', inplace=True)
adata.obs.index.name = None
adata.obs_names_make_unique()

adata.write(obj_save_path + 'adata_obj1.h5ad')

and then loaded this in R using most of the code in the section (https://satijalab.org/seurat/articles/seurat5_bpcells_interaction_vignette.html#parse-biosciences), but I had to make some changes, basically I had to load in the meta data .csv, and ensure it was in a format to be used within the CreateSeuratObject() command. I did do some checks myself to make sure this was being done correctly (which was just checking the order of cell names in the meta data and making sure it matched the order of the cells in the Seurat object)... make sure paths are correct, though, I replaced my path with just ~... note: also ParseBio_PBMC.h5ad is named adata_obj1.h5ad here, but it's the same... :

parse.data <- open_matrix_anndata_hdf5("~/adata_obj1.h5ad")
write_matrix_dir(mat = parse.data, dir = "~/parse_1m_pbmc_example")
parse.mat <- open_matrix_dir(dir = "~/parse_1m_pbmc_example")
parse.meta_data_csv <- read.csv("/cell_metadata_1M_PBMC.csv")
object <- CreateSeuratObject(counts = parse.mat, meta.data = parse.meta_data_csv)

and then from here I continued with the Parse Biosciences Sketch Integration tutorial, here:

object <- NormalizeData(object)
object[["RNA"]] <- split(object[["RNA"]], f = object$sample)
so on and so forth...