Datasets:
license: mit
language:
- en
tags:
- genomics
- yeast
- transcription
- perturbation
- response
- overexpression
pretty_name: Hackett, 2020 Overexpression
size_categories:
- 1M<n<10M
configs:
- config_name: hackett_2020
description: TF overexpression data from Hackett 2020
default: true
dataset_type: annotated_features
metadata_fields:
- regulator_locus_tag
- regulator_symbol
- time
- mechanism
- restriction
- date
- strain
data_files:
- split: train
path: hackett_2020.parquet
dataset_info:
features:
- name: sample_id
dtype: integer
description: >-
unique identifier for a specific sample. The sample ID identifies a
unique (regulator_locus_tag, time, mechanism, restriction, date,
strain) tuple.
- name: db_id
dtype: integer
description: >-
an old unique identifer, for use internally only. Deprecated and
will be removed eventually. Do not use in analysis. db_id = 0, for
GEV and Z3EV, means that those samples are not included in the
original DB.
- name: regulator_locus_tag
dtype: string
description: >-
induced transcriptional regulator systematic ID. See
hf/BrentLab/yeast_genome_resources
- name: regulator_symbol
dtype: string
description: >-
induced transcriptional regulator common name. If no common name
exists, then the `regulator_locus_tag` is used.
- name: target_locus_tag
dtype: string
description: >-
The systematic ID of the feature to which the effect/pvalue is
assigned. See hf/BrentLab/yeast_genome_resources
- name: target_symbol
dtype: string
description: >-
The common name of the feature to which the effect/pvalue is
assigned. If there is no common name, the `target_locus_tag` is
used.
- name: time
dtype: float
description: time point (minutes)
- name: mechanism
dtype:
class_label:
names:
- GEV
- ZEV
description: induction system (GEV or ZEV)
- name: restriction
dtype:
class_label:
names:
- M
- 'N'
- P
description: nutrient limitation (M, N or P)
- name: date
dtype: string
description: date performed
- name: strain
dtype: string
description: strain name
- name: green_median
dtype: float
description: median of green (reference) channel fluorescence
- name: red_median
dtype: float
description: median of red (experimental) channel fluorescence
- name: log2_ratio
dtype: float
description: log2(red / green) subtracting value at time zero
- name: log2_cleaned_ratio
dtype: float
description: Non-specific stress response and prominent outliers removed
- name: log2_noise_model
dtype: float
description: estimated noise standard deviation
- name: log2_cleaned_ratio_zth2d
dtype: float
description: >-
cleaned timecourses hard-thresholded based on multiple observations
(or last observation) passing the noise model
- name: log2_selected_timecourses
dtype: float
description: >-
cleaned timecourses hard-thresholded based on single observations
passing noise model and impulse evaluation of biological feasibility
- name: log2_shrunken_timecourses
dtype: float
description: >-
selected timecourses with observation-level shrinkage based on local
FDR (false discovery rate). Most users of the data will want to use
this column.
Hackett 2020
This Dataset is a parsed version of the data provided by
Calicolabs under the heading "Raw &
processed gene expression data". See scripts/ for more details on the parsing from the
data provided by Calico to this Dataset.
This repo provides 1 dataset:
- hackett_2020: TF overexpression data from Hackett 2020.
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/hackett_2020" dataset from Hugging Face Hub.
from huggingface_hub import ModelCard
from pprint import pprint
card = ModelCard.load("BrentLab/hackett_2020", 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)
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:
from huggingface_hub import snapshot_download
import duckdb
import os
repo_id = "BrentLab/hackett_2020"
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 hackett_2020 parquet file
parquet_path = os.path.join(repo_path, "hackett_2020.parquet")
print(f"✓ Parquet file at: {parquet_path}")
Use your favorite method of interacting with parquet files (eg duckDB, but you could
use dplyr in R or pandas, too).
# Connect to DuckDB and query the parquet file
conn = duckdb.connect()
query = """
SELECT DISTINCT time, mechanism, restriction, date
FROM read_parquet(?)
WHERE regulator_symbol = 'ACA1'
"""
result = conn.execute(query, [parquet_path]).df()
print(f"Found {result}")