simpleaf is a rust framework to make using alevin-fry even simpler. simpleaf encapsulates the process of creating an expanded reference for quantification into a single command (index) and the quantification of a sample into a single command (quant). It also exposes various other functionality, and is actively being developed and expanded.
The simpleaf program can be installed from source, from crates.io, or via bioconda. simpleaf requires, alevin-fry, and either piscem or salmon (or both, if you prefer), as well as wget.
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. If you have any difficulty related to building an index using piscem, before you file an issue on GitHub, please make sure you try to increase your file handle limit (e.g. as is described here).
Check out the detailed documentation here, and read on below to learn more about the background and motivation behind simpleaf.
- Please ensure that the user file handle limit is set to 2048. This may already be set (and should be fine already on OSX), but you can accomplish this by executing:
$ ulimit -n 2048
before running simpleaf.
-
Q(s) : What is the purpose of
simpleaf? Isn't its functionality covered by the constituent programs (e.g.salmon,alevin-fry,piscem, etc.)? Can't I make those tools do the same thingssimpleafdoes? -
A : Yes! It is, of course, possible to replicate the functionality of
simpleafby building a script or workflow around the underlying tools. However,simpleafis designed to make the most common use cases simpler, while also retaining critical flexibility where necessary. Further,simpleafalso provides some extra functionality that one would have to build themselves if wrapping the underlying tools, and simplifies the use of different mapping backends (i.e.salmonorpiscem). For more details, read on below.
The relevant tools that drive simpleaf (i.e. {salmon | piscem} and alevin-fry) are all command-line tools meant to be used together. For those who are very comfortable with the command-line, these tools are designed to be straightforward to use. Further, they are designed to be highly-configurable, so that they can be run in different ways, with different configurations, based upon what the user wants to accomplish. In fact, many users of alevin-fry have crafted their own scripts or pipelines chaining these tools together using bash scripts, custom python scripts, or specially-built pipeline tools like snakemake and nextflow.
While this mode of interaction makes a lot of sense for folks who are very comfortable with the command line and scripting, and who need maximum control over how each aspect of the tools is run, it can seem a bit daunting when one is performing a common task without the need for more exotic configurations. In that case, it should be possible to further simplify the interface to provide a simple command akin to something like cellranger count.
Initially, we designed a Nextflow workflow (quantaf) for wrapping these tools and processing data based on a simple spreadsheet of input. While that approach works well when one needs to process a lot of data, and is easily scalable to many different compute environments thanks to Nextflow, it is a somewhat heavyweight solution. Further, accounting for some current and future directions of development, we also sought a solution where we might selectively employ programmatic (i.e. library-level), rather than file or channel-based communication between the different underlying components.
Therefore, inspired by the flexible yet simple-to-use interface of tools like cellranger (developed by 10X Genomics) and kb-python (developed by the Pachter lab at Caltech), we decided that it made sense to build a stand-alone tool to provide a simplified but flexible interface for our underlying workflows. We also sought to allow some of the modularity provided by tools such as nf-core's scrnaseq pipeline by allowing the use of more than one mapping backend.
While a scripting language like ruby, python or perl is a natural choice for such an intelligent "wrapper" or "pipeline" tool, we chose to develop simpleaf in rust, which actually turns out to work quite well for tasks such as this. While there are several reasons for this decision, a major motivation for this choice is that, as we develop new tools with and transition other functionality over to rust, having simpleaf written in rust will allow for direct programmatic (i.e. library-level) interaction between some of the tools, rather than relying on independent process management and communication.
Finally, while simpleaf is ready-to-use (we use it regularly to process single-cell data), it is still under active development, with new features and capabilities being added. If you have feature suggestions or feedback on directions in which you'd like to see simpleaf grow, please let us know in the issues or discussions.