# Ish: SIMD and GPU accelerated local and semi-global alignment as a CLI filtering tool

**Authors:** Seth Stadick

PMC · DOI: 10.1093/bioadv/vbaf292 · Bioinformatics Advances · 2025-11-17

## TL;DR

Ish is a new command line tool for fast DNA or protein sequence filtering using alignment scores, optimized for performance with SIMD and GPU.

## Contribution

Ish introduces the first index-free, composable CLI tool for alignment-based filtering with high-performance SIMD and GPU support.

## Key findings

- Ish's alignment algorithms achieve or exceed performance of Parasail's reference implementations in terms of GCUPs.
- Ish is open-source and available on GitHub with support for Linux and macOS via Conda.
- The tool allows flexible filtering using various alignment algorithms and record types.

## Abstract

To date, there has been no command line utility for performing index-free alignment-based filtering of records. Since filtering using command line tools is a staple of Bioinformatics, this leaves a gap in command line workflows.

Ish is the first composable unix-style command line tool for filtering the input target records to only those that match the input query with a threshold alignment score, using a selectable alignment algorithm and selectable record type. The core alignment algorithms for ish meet or exceed the performance of their reference implementations in Parasail for both SIMD and GPU alignment, as measured by gigacell updates per second (GCUPs).

The source code and documentation are available at https://github.com/BioRadOpenSource/ish under the Apache-2.0 License and open for community contributions. Ish is installable with Conda and supports Linux and macOS.

## Full-text entities

- **Chemicals:** leucine (MESH:D007930), aspartic acid (MESH:D001224), Blosum62 (-)
- **Mutations:** leucine for isoleucine

## Full text

_Full body text omitted from this summary view._ Fetch the complete paper as Markdown: https://tomesphere.com/paper/PMC12930471/full.md

## Figures

2 figures with captions in the complete paper: https://tomesphere.com/paper/PMC12930471/full.md

## References

10 references — full list in the complete paper: https://tomesphere.com/paper/PMC12930471/full.md

---
Source: https://tomesphere.com/paper/PMC12930471