jo-mengr/cellxgene_pseudo_bulk_full_multiplets_natural_language_annotation

Description

This dataset contains a representation of RNA sequencing data and text descriptions. Dataset type: multiplets (suitable for relevant contrastive-learning or inference tasks).

Cell Sentence Length: The cell sentences in this dataset have a length of $cs_length genes.

The RNA sequencing data used for training was originally gathered and annotated in the CellWhisperer project. It is derived from CellxGene and GEO. Detailed information on the gathering and annotation of the data can be read in the CellWhisperer Manuscript.

Example Data Row

The dataset contains the following column structure (example from the first row):

  sample_idx: census_842c6f5d-4a94-4eef-8510-8c792d1124bc_6077
  cell_sentence_1: census_842c6f5d-4a94-4eef-8510-8c792d1124bc_6077
  cell_sentence_2: MALAT1 MT-CO2 MGP MT-CO1 TPT1 MT-ATP6 FTH1 RPLP1 MT-ND4 S100A6 RPS27A RPS14 RPS8 RPS19 PTMA RPL13A RPS12 RPL19 RPL30 RPL12 RPL11 RPS23 RPL32 RPS3 RPS6...
  positive: This measurement was conducted with 10x 3' v3. A luminal epithelial cell of mammary gland derived from a young African American female donor with an o...
  negative_1_idx: census_842c6f5d-4a94-4eef-8510-8c792d1124bc_6308
  negative_2_idx: census_842c6f5d-4a94-4eef-8510-8c792d1124bc_2346
  adata_link: https://nxc-fredato.imbi.uni-freiburg.de/s/Fa2tMMAz7mAwX4B

The processed .h5ad files used to create this dataset are stored remotely. An example file can be accessed here: https://nxc-fredato.imbi.uni-freiburg.de/s/ic5bF8WoWJnnx45

The AnnData Objects were processed and converted into a Hugging Face dataset using the adata_hf_datasets Python package. The dataset can be used to train a multimodal model, aligning transcriptome and text modalities with the sentence-transformers framework. See mmcontext for examples on how to train such a model.

The anndata objects are stored on nextcloud and a sharelink is provided as part of the dataset to download them. These anndata objects contain intial embeddings generated like this: Each AnnData contained the following embedding keys: ['X_pca', 'X_scvi_fm', 'X_geneformer', 'X_gs10k', 'X_geneformer-v1']. These initial embeddings are used as inputs for downstream model training / inference.

Source

• Original Data: CZ CELLxGENE Discover: A single-cell data platform for scalable exploration, analysis and modeling of aggregated data CZI Single-Cell Biology, et al. bioRxiv 2023.10.30 Publication

  GEO Database: Edgar R, Domrachev M, Lash AE. Gene Expression Omnibus: NCBI gene expression and hybridization array data repository Nucleic Acids Res. 2002 Jan 1;30(1):207-10

• Annotated Data: Cell Whisperer: Multimodal learning of transcriptomes and text enables interactive single-cell RNA-seq data exploration with natural-language chats Moritz Schaefer, Peter Peneder, Daniel Malzl, Mihaela Peycheva, Jake Burton, Anna Hakobyan, Varun Sharma, Thomas Krausgruber, Jörg Menche, Eleni M. Tomazou, Christoph Bock Publication Annotated Data: CellWhisperer website

• Embedding Methods: scVI: Lopez, R., Regier, J., Cole, M.B. et al. Deep generative modeling for single-cell transcriptomics. Nat Methods 15, 1053–1058 (2018). https://doi.org/10.1038/s41592-018-0229-2 geneformer: Theodoris, C.V., Xiao, L., Chopra, A. et al. Transfer learning enables predictions in network biology. Nature 618, 616–624 (2023). Publication

• Further important packages anndata: Isaac Virshup, Sergei Rybakov, Fabian J. Theis, Philipp Angerer, F. Alexander Wolf. anndata: Annotated data. bioRxiv 2021.12.16.473007 Publication scnapy: Wolf, F., Angerer, P. & Theis, F. SCANPY: large-scale single-cell gene expression data analysis. Genome Biol 19, 15 (2018). Publication

Usage

To use this dataset in Python:

from datasets import load_dataset

# Load the dataset
dataset = load_dataset("jo-mengr/cellxgene_pseudo_bulk_full_multiplets_natural_language_annotation_v4")

Understanding the Data Structure

• sample_idx: This column maps to the adata.obs.index of the original AnnData objects
• Chunking: Larger datasets were chunked, so each AnnData object contains only a subset of the indices from the complete dataset
• Share Links: Each row contains a share_link that can be used with requests to download the corresponding AnnData object

Loading AnnData Objects

The share links in the dataset can be used to download the corresponding AnnData objects:

import requests
import anndata as ad

# Get the share link from a dataset row
row = dataset["train"][0]  # First row as example
share_link = row["share_link"]
sample_idx = row["sample_idx"]

# Download and load the AnnData object
response = requests.get(share_link)
if response.status_code == 200:
    with open("adata.h5ad", "wb") as f:
        f.write(response.content)
    adata = ad.read_h5ad("adata.h5ad")

    # The sample_idx corresponds to adata.obs.index
    sample_data = adata[adata.obs.index == sample_idx]
    print(f"Found sample: {sample_data.shape}")
else:
    print("Failed to download AnnData object")