BrentLab/callingcards

Calling Cards

This is data produced in both the Brent Lab and Mitra Lab at Washington University

This repo provides 2 dataset and associated metadata:

• annotated_features: This data scores promoter regions associated with the nearest gene
• genome_map: The binding location data in qbed format

In the annotated features, in order to get the analysis set (you can use duckdb directory instead of tfbpapi -- see the usage section below):

import pandas as pd
from tfbpapi.HfQueryAPI import HfQueryAPI

# Initialize the Hugging Face query API with the calling cards dataset
callingcards_hf = HfQueryAPI(
    repo_id="BrentLab/callingcards", 
    repo_type="dataset"
)

# Set a filter to only include records where data quality passes QC
callingcards_hf.set_filter("annotated_features", data_usable="pass")

# Query all columns from the annotated_features table
# Returns the data as a pandas DataFrame
callingcards_data = callingcards_hf.query(
    "SELECT * FROM annotated_features", 
    "annotated_features"
)

analysis_data = (
    callingcards_data
    .assign(
        # Create a flag: does this regulator have any composite binding?
        has_composite = lambda df: df.groupby('regulator_locus_tag')['composite_binding']
                                      .transform(lambda x: x.notna().any())
    )
    .query(
        # If composite exists for this regulator, require composite to be non-null
        # Otherwise, require single_binding to be non-null
        '(has_composite & composite_binding.notna()) | '
        '(~has_composite & single_binding.notna())'
    )
    .drop(columns=['has_composite'])  # Remove the helper column
)

Usage

The python package tfbpapi provides an interface to this data which eases examining the datasets, field definitions and other operations. You may also download the parquet datasets directly from hugging face by clicking on "Files and Versions", or by using the huggingface_cli and duckdb directly. In both cases, this provides a method of retrieving dataset and field definitions.

tfbpapi

After installing tfbpapi, you can adapt this tutorial in order to explore the contents of this repository.

huggingface_cli/duckdb

You can retrieves and displays the file paths for each configuration of the "BrentLab/callingcards" dataset from Hugging Face Hub.

from huggingface_hub import ModelCard
from pprint import pprint

card = ModelCard.load("BrentLab/callingcards", repo_type="dataset")

# cast to dict
card_dict = card.data.to_dict()

# Get partition information
dataset_paths_dict = {d.get("config_name"): d.get("data_files")[0].get("path") for d in card_dict.get("configs")}

pprint(dataset_paths_dict)

The entire repository is large. It may be preferable to only retrieve specific files or partitions. You can use the metadata files to choose which files to pull.

from huggingface_hub import snapshot_download
import duckdb
import os
# Download only the metadata first
repo_path = snapshot_download(
    repo_id="BrentLab/callingcards",
    repo_type="dataset",
    allow_patterns="annotated_features_meta.parquet"
)

dataset_path = os.path.join(repo_path, "annotated_features_meta.parquet")
conn = duckdb.connect()
meta_res = conn.execute("SELECT * FROM read_parquet(?) LIMIT 10", [dataset_path]).df()
print(meta_res)

We might choose to take a look at the file with id = 1:

# Download only a specific sample's genome coverage data
repo_path = snapshot_download(
    repo_id="BrentLab/callingcards",
    repo_type="dataset",
    allow_patterns="annotated_features/id=1/*.parquet"
)

# Query the specific partition
dataset_path = os.path.join(repo_path, "annotated_features")
result = conn.execute("SELECT * FROM read_parquet(?) LIMIT 10", 
                     [f"{dataset_path}/**/*.parquet"]).df()
print(result)

If you wish to pull the entire repo, due to its size you may need to use an authentication token. If you do not have one, try omitting the token related code below and see if it works. Else, create a token and provide it like so:

repo_id = "BrentLab/callingcards"

hf_token = os.getenv("HF_TOKEN")

# Download entire repo to local directory
repo_path = snapshot_download(
    repo_id=repo_id,
    repo_type="dataset",
    token=hf_token
)

print(f"\n✓ Repository downloaded to: {repo_path}")

# Construct path to the annotated_features_meta parquet file
parquet_path = os.path.join(repo_path, "annotated_features_meta.parquet")
print(f"✓ Parquet file at: {parquet_path}")