Welcome to Portcullis’ documentation

Portcullis stands for PORTable CULLing of Invalid Splice junctions from pre-aligned RNA-seq data. It is known that RNAseq mapping tools generate many invalid junction predictions, particularly in deep datasets with high coverage over splice sites. In order to address this, instead for creating a new RNAseq mapper, with a focus on SJ accuracy we created a tool that takes in a BAM file generated by an RNAseq mapper of the user’s own choice (e.g. Tophat2, Gsnap, STAR2 or HISAT2) as input (i.e. it’s portable). It then, analyses and quantifies all splice junctions in the file before, filtering (culling) those which are unlikely to be genuine. Portcullis output’s junctions in a variety of formats making it suitable for downstream analysis (such as differential splicing analysis and gene modelling) without additional work.

Contents:

• 1. System requirements
• 2. Installation
  • 2.1. From brew
  • 2.2. From bioconda
  • 2.3. From source
• 3. Quickstart
• 4. Using Portcullis
  • 4.1. Prepare
  • 4.2. Junction Analysis
  • 4.3. Junction Filtering
  • 4.4. Bam Filtering
• 5. Metrics
  • 5.1. Junction metrics
  • 5.2. Extra metrics
  • 5.3. Extracted metrics
  • 5.4. Final metrics
• 6. Junctools
  • 6.1. Installation
  • 6.2. Compare
  • 6.3. Convert
  • 6.4. GTF
  • 6.5. Markup
  • 6.6. Set
  • 6.7. Split
• 7. Frequently Asked Questions
  • 7.1. Can I reduce portcullis’ memory usage?

Issues

Should you discover any issues with portcullis, or wish to request a new feature please raise a ticket at https://github.com/maplesond/portcullis/issues. Alternatively, contact Daniel Mapleson at: daniel.mapleson@earlham.ac.uk

Availability and License

Open source code available on github: https://github.com/maplesond/portcullis.git

Spectre is available under GNU GLP V3: http://www.gnu.org/licenses/gpl.txt

Additional Resources

Portcullis was presented at the Genome Science 2016 conference: poster; slides

Authors and Acknowledgements

Daniel Mapleson - Project lead, software developer

David Swarbreck came up with the original plan and has helped design and guide the project from day 1.

Luca Venturini made the logic for the rule-based filtering and has been constantly testing the software helping to find bugs, improve runtime performance at every stage.

Thanks to Gemy Kaithokattil and Shabonham Caim for hunting bugs and providing valuable feedback and Sarah Bastkowski for helping me understanding the various machine learning algorithms and the maths behind it. Also thanks to Chris Bennett for making the logo.

Finally, thanks to the Earlham Insitute and the NBI computing services for supporting the work.