shiny_cuffdiff: On-the-fly interrogation and plotting of gene expression results

[danwiththeplan]

Hi all. I'd like to introduce the community to shiny_cuffdiff

http://nzgl.github.io/shiny_cuffdiff/

Put simply, this code uses the outputs of the Tuxedo workflow (Tophat2 > Cufflinks > Cuffdiff) which is a cuffdiff SQL database file called cuffdiff.db. Then, using the R package cummeRbund, various plots and outputs are created on-the fly.

shiny_cuffdiff consists of two Shiny apps:

gene_expression_panel_cuffdiff:

Takes a cuffdiff database (rebuilding if required) and displays information on a single named gene (as defined by the gene short name or the XLOC number). After the gene information is defined, gene expression plots can be restricted by sample.

gene_comparison_panel_cuffdiff:

Takes a cuffdiff database (rebuilding if required) and displays information on a set of named genes (as defined by the gene short name or the XLOC number). Displays a heatmap and a barplot. After the gene information is defined, gene expression plots can be restricted by sample.

This project is at an early phase of development, but is under active development.

Participants (either users, testers, or code developers) are encouraged!

We are also developing Shiny code to interrogate the outputs of other common gene expression / differential expression workflows (initially starting with DESeq2), which will be published as a separate repository.

Dan Jones
School of Biological Sciences
The University of Auckland

Please feel free to contact me through SeqAnswers or Github.

[Zapages]

Hi

I was just testing the program out. This seems to be a great way to help non-Bioinformaticians to visualize their differential expression data.

There is an extra ")" at the end of the code to run it through shiny in R.

Have you considered something like this: QuickRNASeq http://sourceforge.net/projects/quickrnaseq/ or Degust: http://vicbioinformatics.com/degust/...?code=example#

[danwiththeplan]

Hi, thanks for your reply. I'm not able to replicate this error. If you mean this error:

error in evaluating the argument 'object' in selecting a method for function 'expressionPlot': Error in sqliteSendQuery(con, statement, bind.data) : 
  error in statement: near ")": syntax error

..then this occurs when you aren't giving the program a valid cuffdiff directory (not the cuffdiff.db file, the cuffdiff directory), or the gene name or XLOC number does not exist. Note that the plotting isn't instant, if you can see cummeRbund getting the gene information:

Getting gene information:
	FPKM
	Differential Expression Data
	Annotation Data
	Replicate FPKMs
	Counts

.then it's working and the plots will pop up soon.

This is not so much an error in the code, more an example of error handling that could be more clear and informative. As is noted, this is a program in early development and I do note that error handling is the next thing I'm working on. Additionally, the plots are not that nice, so I will be using ggplot2 to make nicer plots in the very near future.

I'm aware of Degust and QuickRNAseq. Degust does something very similar but does not handle cuffdiff outputs at all. Valid comparison, but until they add capability to handle cuffdiff, I think this work has a place. Additionally Degust requires that you upload your data (EDIT: you can download the code and run it on your computer, so I'm wrong here)
Not sure the comparison to QuickRNAseq is valid, since it's an entire workflow that starts right from the FASTQ file (as I understand it).

[lgoff]

Really interesting project. Looking forward to contributing to it's development. Thanks for picking up this mantle.

-Loyal

[Zapages]

Yes, I agree this work definitely is useful and has a place. As this could be specifically designed for cuffdiff and other R based tool sets.

Currently, using gplot2 to create heatmaps takes a bit of time with R for DeSeq and EdgeR data.

I was able to fix the error, by going into the load_packages.R files and fixing up the URL addresses. My R was not accepting the 's' in https://...

I look forward to future versions of your tool.

Best Regards.

[danwiththeplan]

Yes, I agree this work definitely is useful and has a place. As this could be specifically designed for cuffdiff and other R based tool sets.

Currently, using gplot2 to create heatmaps takes a bit of time with R for DeSeq and EdgeR data.

I was able to fix the error, by going into the load_packages.R files and fixing up the URL addresses. My R was not accepting the 's' in https://...

I look forward to future versions of your tool.

Best Regards.

Great, thanks, this is now fixed. Prettier plots (using ggplot2) are definitely a plan for the near future. Hoping to get publication-level plots that can be generated on the fly. To avoid spamming Seqanswers, it might be better to keep bug reports as personal messages or in github.

Questions and clarifications on how to use the software are fine

[Zapages]

Hi

I just tested the program out again with all of my RNA-Seq datasets. It works really great. Thank you for creating this application.

Maybe to help users create publication like graphs/charts, maybe there could be an option to save the images as *.svg files. This could allow the user to edit svg file (with svg editors) and have publication like results.

Also if I were to use this for publication. How should we cite the application?

Many thanks in advance.

-Zain

[danwiththeplan]

Hi

I just tested the program out again with all of my RNA-Seq datasets. It works really great. Thank you for creating this application.

Maybe to help users create publication like graphs/charts, maybe there could be an option to save the images as *.svg files. This could allow the user to edit svg file (with svg editors) and have publication like results.

Also if I were to use this for publication. How should we cite the application?

Many thanks in advance.

-Zain

Thanks for the feedback! To answer your questions:

1) For future releases, we will include citation information, but for now, please cite:

Jones,D., Stuckey, A., and Fan, V. (2016). shiny_cuffdiff (version 0.1): On-the-fly interrogation and plotting of gene expression results. http://nzgl.github.io/shiny_cuffdiff/

Remember to also cite the underlying R code (cummeRbund) and the Tuxedo workflow.

2) Excellent suggestion regarding .svg files. I will add this to the list of features we are implementing. The ability to save graphics in various formats is definitely on the to-do list.

Following the project on Github is a good idea, as we are actively working on new features and better error handling. For example, we have already implemented better methods of selecting the database file and are working on making the graphics more visually pleasing and customisable. We expect to release version 0.2 early next week, so keep an eye out!

Again, thanks for the feedback, please Star us on Github!

[MikhailFokin]

nice a bit scary to have it in a browser...
what the "include replicates" option means?