bricks2marble

• Python structures for nucleotide sequences and genome annotations.
• Tensorflow implementation of an HMM used for finding genes.
• Pre- and postprocessing tools for deep learning genome annotation models.
• Python interfaces for common bioinformatics tools and file format converters.

Installation

Download bricks2marble via pip.

$ python -m pip install bricks2marble

For development purposes, clone the repository and install it locally inside your virtual environment.

$ git clone https://github.com/gaius-augustus/bricks2marble
$ cd bricks2marble
$ python -m pip install -e .

If access to the tensorflow part of bricks2marble is needed, specify this as an optional dependency when installing. This will install hidten.

$ python -m pip install bricks2marble[tf]
# or
$ python -m pip install -e .[tf]

When plotting is required, pip install bricks2marble[plot].

Overview

Below are some use cases of bricks2marble. All methods have docstrings that explain their behaviour and several optional arguments in detail.

Reading and writing

Loading large fasta files is implemented efficiently using bytearray translation tables (~11 seconds for the human genome). Additionally, mmap is used for indexing large fasta files (~4 seconds for the human genome).

import bricks2marble as b2m

fasta = b2m.io.load_fasta("genome.fa.gz") # load everything into memory
sequence = fasta["chr1"].positions(0, 100)

fasta = b2m.io.indexed_fasta("genome.fa") # build a sequence index
sequence = fasta.fetch("chr1", (0, 100)) # load only required parts

For some specific cases, external tools are used. For example, indexing .fa.gz files requires pyfaidx.

$ python -m pip install pyfaidx

fasta = b2m.io.indexed_fasta("genome.fa.gz") # build a sequence index for compressed files
sequence = fasta.fetch("chr1", (0, 100)) # load only required parts

Additionally, .gp (.genepred) files can be loaded and are internally sorted for optimized access.

anno = b2m.io.load_annotation("reference.gp")
anno.classify(1062, "chr1") # labels per strand: ("intergenic", "CDS")

Tools

The subpackage bricks2marble.tools contains a number of interfaces to common external tools related to genome file formats. Download the external tools yourself and tell bricks2marble where they can be found locally. Optionally, you can add them to your system path, so bricks2marble can find them automatically.

Example: Comparing genome annotations

Download gffcompare and use the bricks2marble interface for extracting metrics.

import bricks2marble as b2m

b2m.tools.configure(gffcompare="path/to/gffcompare")
comparison = b2m.tools.compare(
    ["my_annotation.gp", "other_annotation.gtf"],
    "reference.gff",
    e=3,
)
print(comparison[0].locus.sensitivity)
fig = b2m.tools.plot.comparison(
    comparison,
    labels=["My", "Other"],
    table=True,
)
fig.show()

Example: Converting files

Convert various file formats for genome annotations. The internal bricks2marble representation of annotation files is closely related to the genepred format. Conversions to gtf and gff3 are implemented directly. Conversions from these formats to genepred are handled by the corresponding external tools from UCSC, like gtfToGenePred.

import bricks2marble as b2m

b2m.tools.configure(gtfToGenePred="path/to/gtfToGenePred")

with b2m.tools.Converter("my_annotation.gtf", "gp") as tmp_file_path:
    # gp file created using Python's tempfile
    annotation = b2m.io.load_annotation(tmp_file_path)
# gp file deleted, annotation loaded into memory

b2m.tools.convert(annotation, "my_annotation.gff", source="MyTool")

License

This project is licensed under the MIT license.

Projects using bricks2marble

• Tiberius: End-to-End Deep Learning with an HMM for Gene Prediction