Note: this code is not yet fully functional! There are bugs that I am still working out. I will remove this statement and tag a release when the code is entirely debugged. That said, the code is essentially complete, with full implementation of the basic pipeline.
This project started out as a hack of PBJelly in an attempt to improve the efficiency of the code. I ended up rebuilding the core pipeline, focusing it down into four stages: Setup, Support, Assembly, and Placement. The code is designed for compatibility with the most recently developed standards, requiring PacBio subreads in the BAM format. It makes use of the cutting edge algorithms Minimap, Miniasm, and Racon, for ultra-fast assembly of gap sequences. The stages are managed by a single driver script, Jelly2.py, so this program is easy to use. To see the full range of options, run python Jelly2.py --help, but the basic usage is:
$ python Jelly2.py [options] <Scaffolds.fa> <Subreads.bam>
Make sure that the src directory is in your PYTHONPATH variable, so the driver script can find the modules. Though this repository was created on May 8, 2017, it contains the git tracking history of the above-linked PBJelly repository in which this project began.
I recommend using Pitchfork to deploy BLASR, along with SAMtools and bam2fastx
To find PacBio reads that span the gaps predicted in the scaffolds, gap-flanking sequences are extracted from the assembly. Separate Fasta files are created for the left flanks and the right flanks. The PacBio reads are then aligned against each flank file with BLASR. Supporting reads are those that align to each flank, spanning the predicted gap size within a certain margin of error. If there are enough support reads for a gap, the reads will be assembled with Minimap and Miniasm. A consensus will be determined with Minimap and Racon. If the assembled gap sequence passes quality control, it will be placed into the scaffold, filling the gap.
Any questions can be addressed to: dbrowne@tamu.edu