Tool:MetaR: a simple language useful for RNA-Seq data analysis

Question

[fac2003]

Hello,

I wanted to introduce MetaR, a language to analyze RNA-Seq data and produce visualizations such as heatmaps. MetaR provides a language simpler to learn and use than R, yet generates full analysis scripts in R, using popular R/Bioconductor packagages (e.g., data.table, EdgeR, pheatmap). We find that on average 1 line of MetaR generates about 5-10 lines of R code. The language offers full auto-completion for table, columns, model or plot names, making it easier for beginners to get started.

At the end of this post is a snapshot of an RNA-Seq analysis followed by construction of a heatmap with the MetaR language.

This blog post has more details about motivations and design and links to a video demonstrating how to build a heatmap from a table of read counts with MetaR. For the technically inclined, MetaR is a composable language, which means that it is easy to extend with new instructions, expressions, or other language constructs, in ways that are not practical with traditional languages. You can even extend the language as you are writing a new script and realize you need a new kind of statement.

MetaR is open-source (Apache 2.0).

[Irsan]

To be honest I think this looks too difficult for a biologist and too not flexible for a bio-informatician. The future is a server oriented omics complete workbench with a nice web front end.

[fac2003]

We are starting to teach it to our local investigators (a mix of biologists and clinician scientists). So far so good. If you don't have somebody nearby to show you I realize it can be more challenging. We'll have to write good documentation moving forward, and that is certainly a challenge for any new tool or language.

As for flexibility, it is more flexible than any technology are probably familiar with because any user with programming experience can extend the language.

See when I add the edgeR language in the first video as an example of this type of flexibility. A presentation about language composition and the platform this is based on can be found here.

[pld]

I think this is a great idea, especially if it is expanded to multiple Bioconductor packages (e.g. Limma). I think there are many people who may not be bioinformatics by training or very familiar with programming but have a desire to approach them and are turned off or get stuck when they run into the byzantine nature of R.

With the right documentation I think this tool could really increase the accessibility of bioinformatic analysis.

[fac2003]

Thanks. Yes, Limma is on our Roadmap. We will start with Limma Voom, as it appears to have done well in benchmarks for RNA-Seq.

Regarding documentation, we will write a comprehensive doc after we have more experience with the problems our local users encounter. We are still tweaking the tool from feedback we are receiving. At this time, the videos are the only documentation available, plus the first chapters of the MPS book, which explain the UI of the MPS platform. The written doc will be written in the same style as the book (but will be free to download).

[fac2003]

An update:

• we have posted a documentation booklet (see Documentation section of the web site for a PDF download).
• MetaR now also runs on Windows.

[fac2003]

We've added a comprehensive manual, available for tablets on Google Play or as a PDF.

Answer 1

[fac2003]

I have updated the snapshot to avoid the initial impression that some people could have: MetaR is not a text-based language. Users do not need to know some complex syntax before they can start writing analyses. Instead, they can use auto-completion to write analyses and configure them. The videos should illustrate this nicely.

Answer 2

[fac2003]

An update: Version 1.5 of MetaR supports the full R language. This R language is composable and can be extended with micro-languages, as illustrated in the following example.

More details are available in the "Composable R" Chapter of the documentation.