[Announcement] Testers needed for new peak annotation tool

[amitra]

Hi,

I am developing a new c++ based open-source tool (peak-tool) to annotate peaks from a bed file using gencode annotation database.

Once the gencode database is loaded and processed, each bed line lookup takes a constant time because of an indexing data structure.

Here are example annotation:

Bed:
chr14 57735721 57735722 MACS_peak_13241 1159.23

Annotated:
chr14 57735721 MACS_peak_13241 1159.23 PROMOTER AP5M1 + AP5M1-001 protein_coding 57735627 94
chr14 57735721 MACS_peak_13241 1159.23 PROMOTER EXOC5 - EXOC5-001 protein_coding 57735726 5

Bed:
chr1 12556659 12556660 MACS_peak_330 1733.05

Annotated:
chr1 12556659 MACS_peak_330 1733.05 PROMOTER VPS13D + VPS13D-012 retained_intron 12557280 -621


Bed:
chr1 1778750 1778751 MACS_peak_51 102.12

Annotated:
chr1 1778750 MACS_peak_51 102.12 INTRON GNB1 - GNB1-001 protein_coding 1822495 43745



Can you please test this tool and provide feedback for further improvements?

Here's the link to github
https://github.com/goxed/peak-tool

Right now the tool can annotate only from Bed files (human / hg19) using annotations from Gencode database (included in the git repo)

The tool needs 16GB RAM on MAC OS X (10.9.x or greater) systems and >=20GB RAM on Linux systems (16GB if you use ZRAM memory compression or a very fast SSD swap)

[dpryan]

What's the benefit of this over things like bedtools or bedops? Also, it looks like you're actually annotating the midpoint of peaks, rather than peaks themselves (e.g., https://github.com/goxed/peak-tool/b....cpp#L496-L498). That would seem rather problematic.

[amitra]

What's the benefit of this over things like bedtools or bedops? Also, it looks like you're actually annotating the midpoint of peaks, rather than peaks themselves (e.g., https://github.com/goxed/peak-tool/b....cpp#L496-L498). That would seem rather problematic.

Thanks for the feedback!

I would imagine this to be a tool specifically for converting peaks to gene names. (currently peak summits, but I plan on adding more features including coverage in future, to address the issue that you mention).

The tool uses an in-memory index of the gencode features, to make the search extremely fast, and is suitable for large bed files.