PALSS

PALSS (Pangenome Graph Augmentation from Long-reads Specific Strings) is an assembly- and mapping-free approach for updating (or augmenting) a pangenome graph directly from unassembled long reads sequenced from a new individual not already in the pangenome.

Installation

PALSS has been tested only on 64bit Linux system(s).

git clone https://github.com/ldenti/palss
cd palss ; mkdir build ; cd build
cmake ..
make -j2
cd ..
./palss -h

Usage guide

PALSS starts from a pangenome graph (.gfa) and a read sample (.fx, can be gzipped) and produces the corresponding augmented pangenome graph (.gfa).

We explain how to use PALSS using the example data available in the example subdirectory.

Note: we suggest to run PALSS on error-corrected reads and on small- to medium-sized pangenome graphs.

# get paths from graph (assuming vg to be in $PATH)
vg paths -F -x example/reference.gfa > example/reference.paths.fa

# build FMD-index from paths of the graph
./build/rb3-prefix/src/rb3/ropebwt3 build -d example/reference.paths.fa > example/reference.paths.fa.fmd

# sketch the graph using 27-mers solid anchors
# !!! Set the number of paths (haplotypes) in the graph using the -g option !!!
./palss sketch -g1 -k27 example/reference.gfa example/reference.paths.fa.fmd > example/reference-k27.skt

# search for specific strings in the haplotypes
./palss search -k27 example/reference.gfa example/reference-k27.skt example/reference.paths.fa.fmd example/reads.fa > example/sfs.txt

# cluster specific strings and analyze clusters
./palss call -k27 example/reference.gfa example/reference-k27.skt example/sfs.txt example/reads.fa > example/new_portions.gaf

# augment the graph
vg augment --min-coverage 1 --gaf example/reference.gfa example/new_portions.gaf > example/reference-augmented.gfa

Solid anchors analysis

To analyze solid anchors from a pangenome wrt any fastx file (use -r if .fq):

./palss kan [-r] [.skt] [.fx] > [.bed]

Please refer to the scripts in the exps subdirectory for more information.

Experiments

Instructions and code to reproduce the experiments described in the manuscript can be found here.

Planned future improvements

• cyclic graphs
• align consensus directly to subgraphs
• works from .vg/.gbwt
• parallelize, improve scalability

For any question/doubt, please open an issue.