alevin-fry is a suite of tools for the rapid, accurate and memory-frugal processing single-cell and single-nucleus sequencing data. It consumes RAD files generated by piscem or salmon alevin, and performs common operations like generating permit lists, and estimating the number of distinct molecules from each gene within each cell. The focus in alevin-fry is on safety, accuracy and efficiency (in terms of both time and memory usage).

You can read the paper describing alevin fry, "Alevin-fry unlocks rapid, accurate, and memory-frugal quantification of single-cell RNA-seq data" here, and the pre-print on bioRxiv.

Note: We recommend using piscem as the back-end mapper, rather than salmon, as it is substantially more resource-frugal, faster, and is a larger focus of current and future development.

While this README contains some useful information to get started and some pointers, alevin-fry has it's own dedicated documentation site, hosted on ReadTheDocs.

• Quickstart guide using the simpleaf wrapper

• Relationship to alevin: Alevin-fry has been designed as the successor to alevin. It subsumes the core features of alevin, while also providing important new capabilities and considerably improving the performance profile. We anticipate that new method development and feature additions will take place primarily within the alevin-fry codebase. Thus, we encourage users of alevin to migrate to alevin-fry when feasible. That being said, alevin is still actively-maintained and supported, so if you are using it and not ready to migrate you can continue to ask questions and post issues in the salmon repository.

Are you curious about processing details like whether to use a sparse or dense index? Do you have a question that isn't necessarily a bug report or feature request, and that isn't readily answered by the documentation or tutorials? Then please feel free to ask over in the Q&A.

• The generation of the reduced alignment data (RAD) files processed by alevin-fry is done by either piscem or salmon. The latest version of both are available on GitHub and via bioconda.

• The simpleaf repository contains a dedicated wrapper / workflow runner for processing data with alevin-fry that vastly simplifies both the creation of extended references and the subsequent quantification of samples. If you find that simpleaf is missing a feature that you'd like to have, please consider submitting a feature request in the simpleaf repository.

• The pyroe repository provides tools to help easily construct an enhanced (spliced + intronic or spliced + unspliced) transcriptome from a reference genome and GTF file.

• The fishpond package — maintained by @mikelove and his lab — contains the recommended relevant functions for reading alevin-fry output (particularly USA-mode output) into the R ecosystem, in the form of a singleCellExperiment object.

• The alevinqc package — maintained by @csoneson — provides tool and functions for performing quality control and assessment downstream of alevin-fry.

Alevin-fry is available for both x86 linux and OSX platforms using bioconda. On Apple silicon, you can either build (easily) from source (see below) or run alevin-fry under the rosetta 2 emulation layer.

With bioconda in the appropriate place in your channel list, you should simply be able to install via:

$ conda install -c bioconda alevin-fry

Alevin-fry can also be installed from crates.io using cargo. This can be done with the following command:

$ cargo install alevin-fry

If you want to use features or fixes that may only be available in the latest develop branch (or want to build for a different architecture), then you have to build from source. Luckily, cargo makes that easy; see below.

Alevin-fry is built and tested with the latest (major & minor) stable version of Rust. While it will likely compile fine with slightly older versions of Rust, this is not a guarantee and is not a support priority. Unlike with C++, Rust has a frequent and stable release cadence, is designed to be installed and updated from user space, and is easy to keep up to date with rustup. Thanks to cargo, building should be as easy as:

$ cargo build --release

subsequent commands below will assume that the executable is in your path. Temporarily, this can be done (in bash-like shells) using:

$ export PATH=`pwd`/target/release/:$PATH

If you use alevin-fry in your work, please cite:

He, D., Zakeri, M., Sarkar, H., Soneson, C., Srivastava, A., and Patro, R. Alevin-fry unlocks rapid, accurate and memory-frugal quantification of single-cell RNA-seq data. Nat Methods 19, 316–322 (2022). https://doi.org/10.1038/s41592-022-01408-3

BibTeX:

@Article{He2022,
author={He, Dongze and Zakeri, Mohsen and Sarkar, Hirak and Soneson, Charlotte and Srivastava, Avi and Patro, Rob},
title={Alevin-fry unlocks rapid, accurate and memory-frugal quantification of single-cell RNA-seq data},
journal={Nature Methods},
year={2022},
month={Mar},
day={01},
volume={19},
number={3},
pages={316-322},
issn={1548-7105},
doi={10.1038/s41592-022-01408-3},
url={https://doi.org/10.1038/s41592-022-01408-3}
}