GenomicArrays 0.0.1

Genomic region based arrays backed by TileDB

Genomic Arrays based on TileDB

GenomicArrays is a Python package for converting genomic data from BigWig format to TileDB arrays.

Installation

Install the package from PyPI

pip install genomicarrays

Quick Start

Build a GenomicArray

Building a GenomicArray generates 3 TileDB files in the specified output directory:

• feature_annotation: A TileDB file containing input feature intervals.
• sample_metadata: A TileDB file containing sample metadata, each BigWig file is considered a sample.
• A matrix TileDB file named by the layer_matrix_name parameter. This allows the package to store multiple different matrices, e.g. 'coverage', 'some_computed_statistic', for the same interval, and sample metadata attributes.

The organization is inspired by the SummarizedExperiment data structure. The TileDB matrix file is stored in a features X samples orientation.

To build a GenomicArray from a collection of BigWig files:

import numpy as np
import tempfile
import genomicarrays as garr

# Create a temporary directory, this is where the
# output files are created. Pick your location here.
tempdir = tempfile.mkdtemp()

# List BigWig paths
bw_dir = "your/biwig/dir"
files = os.listdir(bw_dir)
bw_files = [f"{bw_dir}/{f}" for f in files]

features = pd.DataFrame({
     "chrom": ["chr1", "chr1"],
     "start": [1000, 2000],
     "end": [1500, 2500]
})

# Build GenomicArray
garr.build_genomicarray(
     files=bw_files,
     output_path=tempdir,
     features=features,
     # agg function to summarize mutiple values
     # from bigwig within an input feature interval.
     feature_annotation_options=garr.FeatureAnnotationOptions(
        aggregate_function = np.nanmean
     ),
     # for parallel processing multiple bigwig files
     num_threads=4
)

The build process stores missing intervals from a bigwig file as np.nan. The default is to choose an aggregate functions that works with np.nan.

Note

This project has been set up using PyScaffold 4.6. For details and usage information on PyScaffold see https://pyscaffold.org/.