biobear 0.20.2

A package for working with Bioinformatics data with SQL and Arrow

biobear is a Python library designed for reading and searching bioinformatic file formats, using Rust as its backend and producing Arrow Batch Readers and other downstream formats (like polars or duckdb).

The python package has minimal dependencies and only requires Polars. Biobear can be used to read various bioinformatic file formats, including FASTA, FASTQ, VCF, BAM, and GFF locally or from an object store like S3. It can also query some indexed file formats locally like VCF and BAM.

Please see the documentation for information on how to get started using biobear.

Quickstart

To install biobear, run:

pip install biobear
pip install polars # needed for `to_polars` method

Create a file with some GFF data:

echo "chr1\t.\tgene\t1\t100\t.\t+\t.\tgene_id=1;gene_name=foo" > test.gff
echo "chr1\t.\tgene\t200\t300\t.\t+\t.\tgene_id=2;gene_name=bar" >> test.gff

Then you can use biobear to read a file:

import biobear as bb

session = bb.connect()
df = session.sql("""
    SELECT * FROM gff_scan('test.gff')
""").to_polars()

print(df)

This will print:

┌─────────┬────────┬──────┬───────┬───┬───────┬────────┬───────┬───────────────────────────────────┐
│ seqname ┆ source ┆ type ┆ start ┆ … ┆ score ┆ strand ┆ phase ┆ attributes                        │
│ ---     ┆ ---    ┆ ---  ┆ ---   ┆   ┆ ---   ┆ ---    ┆ ---   ┆ ---                               │
│ str     ┆ str    ┆ str  ┆ i64   ┆   ┆ f32   ┆ str    ┆ str   ┆ list[struct[2]]                   │
╞═════════╪════════╪══════╪═══════╪═══╪═══════╪════════╪═══════╪═══════════════════════════════════╡
│ chr1    ┆ .      ┆ gene ┆ 1     ┆ … ┆ null  ┆ +      ┆ null  ┆ [{"gene_id","1"}, {"gene_name","… │
│ chr1    ┆ .      ┆ gene ┆ 200   ┆ … ┆ null  ┆ +      ┆ null  ┆ [{"gene_id","2"}, {"gene_name","… │
└─────────┴────────┴──────┴───────┴───┴───────┴────────┴───────┴───────────────────────────────────┘

Using a Session w/ Exon

BioBear exposes a session object that can be used with exon to work with files directly in SQL, then eventually convert them to a DataFrame if needed.

See the BioBear Docs for more information, but in short, you can use the session like this:

import biobear as bb

session = bb.connect()

session.sql("""
CREATE EXTERNAL TABLE gene_annotations_s3 STORED AS GFF LOCATION 's3://BUCKET/TenflaDSM28944/IMG_Data/Ga0451106_prodigal.gff'
""")

df = session.sql("""
    SELECT * FROM gene_annotations_s3 WHERE score > 50
""").to_polars()
df.head()
# shape: (5, 9)
# ┌──────────────┬─────────────────┬──────┬───────┬───┬────────────┬────────┬───────┬───────────────────────────────────┐
# │ seqname      ┆ source          ┆ type ┆ start ┆ … ┆ score      ┆ strand ┆ phase ┆ attributes                        │
# │ ---          ┆ ---             ┆ ---  ┆ ---   ┆   ┆ ---        ┆ ---    ┆ ---   ┆ ---                               │
# │ str          ┆ str             ┆ str  ┆ i64   ┆   ┆ f32        ┆ str    ┆ str   ┆ list[struct[2]]                   │
# ╞══════════════╪═════════════════╪══════╪═══════╪═══╪════════════╪════════╪═══════╪═══════════════════════════════════╡
# │ Ga0451106_01 ┆ Prodigal v2.6.3 ┆ CDS  ┆ 2     ┆ … ┆ 54.5       ┆ -      ┆ 0     ┆ [{"ID",["Ga0451106_01_2_238"]}, … │
# │ Ga0451106_01 ┆ Prodigal v2.6.3 ┆ CDS  ┆ 228   ┆ … ┆ 114.0      ┆ -      ┆ 0     ┆ [{"ID",["Ga0451106_01_228_941"]}… │
# │ Ga0451106_01 ┆ Prodigal v2.6.3 ┆ CDS  ┆ 1097  ┆ … ┆ 224.399994 ┆ +      ┆ 0     ┆ [{"ID",["Ga0451106_01_1097_2257"… │
# │ Ga0451106_01 ┆ Prodigal v2.6.3 ┆ CDS  ┆ 2261  ┆ … ┆ 237.699997 ┆ +      ┆ 0     ┆ [{"ID",["Ga0451106_01_2261_3787"… │
# │ Ga0451106_01 ┆ Prodigal v2.6.3 ┆ CDS  ┆ 3784  ┆ … ┆ 114.400002 ┆ +      ┆ 0     ┆ [{"ID",["Ga0451106_01_3784_4548"… │
# └──────────────┴─────────────────┴──────┴───────┴───┴────────────┴────────┴───────┴───────────────────────────────────┘

Ecosystem

BioBear aims to make it simple to move easily to and from different prominent data tools in Python. Generally, if the tool can read Arrow, it can read BioBear's output. To call out a few examples here:

Polars

The session results and Reader objects can be converted to a Polars DataFrame.

import biobear as bb

session = bb.connect()

df = session.sql("""
    SELECT * FROM gff_scan('test.gff')
""").to_polars()

Known Issues

For GenBank and mzML, the naive SELECT * will cause an error, because Polars doesn't support all Arrow types -- Map being the specific offender here. In these cases, select the fields from the map individually. Alternatively, you can first convert the table to a Pandas DataFrame.

DuckDB

BioBear can also be used to read files into a duckdb database.

import biobear as bb
import duckdb

session = bb.connect()

session.sql("""
    CREATE EXTERNAL TABLE gene_annotations STORED AS GFF LOCATION 'python/tests/data/test.gff'
""")

result = session.sql("""
    SELECT * FROM gene_annotations
""")

gff_table_arrow_table = result.to_arrow()

duckdb_conn = duckdb.connect()

result = duckdb_conn.execute('SELECT * FROM gff_table_arrow_table').fetchall()
print(result)

Performance

Please see the exon's performance metrics for thorough benchmarks, but in short, biobear is generally faster than other Python libraries for reading bioinformatic file formats.

For example, here's quick benchmarks for reading one FASTA file with 1 million records and reading 5 FASTA files each with 1 million records for the local file system on an M1 MacBook Pro:

Library	1 file (s)	5 files (s)
BioBear	4.605 s ± 0.166 s	6.420 s ± 0.113 s
BioPython	6.654 s ± 0.003 s	34.254 s ± 0.053 s

The larger difference multiple files is due to biobear's ability to read multiple files in parallel.