Source-linked AI summary
Analysing high-throughput sequencing data in Python with HTSeq 2.0
Givanna H Putri, Simon Anders, Paul Theodor Pyl, John E Pimanda, Fabio Zanini
TL;DR
HTSeq was initially designed as a general-purpose sequencing tool, with gene counting aimed at bulk RNA-seq, while single-cell omics grew and many experimental approaches relied on open-source software. HTSeq 2.0 expands the API and updates counting and sparse-data support, and is presented as a fast, reliable library for high-throughput sequencing analysis and bulk and single-cell RNA-seq quantification.
Problem
HTSeq was initially a general-purpose sequencing tool with bulk-focused counting, while single-cell omics were growing and many experimental approaches relied on open-source software.
Source: Text · pp. 1–2Method
HTSeq 2.0 expands its API, updates htseq-count for single-cell workflows, adds an islands-of-data sparse representation, and modernizes documentation, testing, deployment, and Python support.
Source: Text · pp. 2–3Results
HTSeq 2.0 is presented as a fast and reliable Python library for high-throughput sequencing analysis and bulk and single-cell RNA-seq quantification.
Source: Text · p. 3Takeaways & Limitations
The authors conclude that HTSeq 2.0 is a convenient tool for exploring and quantifying high-throughput sequencing results across multiple omic modalities.
Source: Text · p. 4
Abstract
from arXiv · showhide
Summary: HTSeq 2.0 provides a more extensive API including a new representation for sparse genomic data, enhancements in htseq-count to suit single cell omics, a new script for data using cell and molecular barcodes, improved documentation, testing and deployment, bug fixes, and Python 3 support. Availability and implementation: HTSeq 2.0 is released as an open-source software under the GNU General Public Licence and available from the Python Package Index at https://pypi.python.org/pypi/HTSeq. The source code is available on Github at https://github.com/htseq/htseq. Contact: fabio.zanini@unsw.edu.au
Text
HTSeq 2.0 extends its Python API and counting tools for bulk and single-cell sequencing, while adding sparse genomic data support and improvements to documentation, testing, and deployment.
Text
- HTSeq 2.0 extends htseq-count and the general-purpose API to support diverse omics analyses, including bulk and single-cell RNA-Seq.The update was designed to encompass analyses beyond the script’s original bulk RNA-Seq focus.Source: Text · pp. 2–3
- Multiple BAM files can be counted in one call, processed in parallel across cores, and written in formats including sparse matrices, mtx, h5ad, and loom.Feature metadata can be retained, including chromosome or alias information, and exon-level counts can be disaggregated to quantify differential isoform expression.Source: Text · p. 2
- HTSeq adds htseq-count-barcodes to quantify features in data multiplexed by cell barcodes and UMIs, including custom re-analysis of cellranger BAM files.The script allows cellranger-produced BAM files to be re-analysed with different parameters.Source: Text · p. 2
- StretchVector represents genomic “islands of knowledge” as dense numpy-array stretches with coordinates, leaving the rest of the genome unstored.Its operations include extension, trimming, resetting, shifting, slicing, copying, and conversion to and from monolithic arrays; HTSeq also adds bedGraph and BigWig parsers and tutorials for custom ChIP-Seq and chromatin-conformation analyses.Source: Text · p. 3
- HTSeq modernizes its codebase for Python 3 and adds continuous integration, binary releases, tests, bug fixes, and HTSlib-based SAM/BAM/CRAM autodetection.The authors report that these changes preserve HTSeq’s efficiency, supported by its Cython and SWIG design.Source: Text · p. 3
Figure
Figure 1 summarizes HTSeq 2.0 improvements to htseq-count and sparse genomic data representations. It depicts multicore processing and output options for single-cell data, gene- and exon-level quantification, and two sparse-data structures.
Figure
- Parallel processing across multicore architectures enables faster processing of single-cell data represented by one BAM file per cell, with new output formats available.The figure notes that one-BAM-per-cell data is typical for Smart-seq2 and viscRNA-Seq.Source: Figure 1 · p. 4
- The conventional gene-cell matrix collapses reads aligning to distinct exons of the same gene into one gene count, while additional attributes support exon-level quantification with gene identity retained.These are complementary gene- and exon-level representations in htseq-count.Source: Figure 1 · p. 4
- HTSeq 2.0 represents sparse genomic data with StepVector for piecewise-constant data and StretchVector for sparse islands.Source: Figure 1 · p. 4