#bioinformatics #long-read

app fpa_lr

fpa filter long read mapping information to save disk space

6 releases (breaking)

0.5.0 May 14, 2019
0.4.0 Feb 14, 2019
0.3.0 Oct 16, 2018
0.2.0 Oct 12, 2018
0.1.1 Oct 9, 2018

#11 in Science

36 downloads per month

MIT license


fpa Filter Pairwise Alignment

Build Status

Filter output of all-against-all read mapping, you filter or select:

  • internal match
  • containment
  • dovetails
  • self matching
  • read name match against regex
  • length of overlap
  • length of read in overlap

For internal match, containment, dovetails definition go read algorithm 5 in minimap article


Long Read mapping tools provides all match they found in read dataset, for many usage some of match aren't usfull, this programme provide some filter to remove it. This soft can be replace by a simple awk, bash, python, perl, {your favorite language}.


fpa -i <input> -o <output> <option> <subcommand: drop | keep | index | rename | gfa>

Subcommand can be split in two group:

  • filters (drop, keep), select wich overlap are write in output
  • generators (index, rename, gfa), generate new data from overlap

By default input and output are stdin and stdout so you can use like this:

minimap2 long_read.fasta long_read.fasta | fpa keep -d | gzip - > only_dovetail.paf.gz
minimap2 long_read.fasta long_read.fasta | fpa drop -l 500 -L 2000 > only_between_500_2000.paf
minimap2 long_read.fasta long_read.fasta | fpa drop -s -m read_1 > no_self_no_match_read_1.paf
minimap2 long_read.fasta long_read.fasta | fpa drop -s rename -o rename.csv > no_self_match_renamed.paf
minimap2 long_read.fasta long_read.fasta | fpa drop -s rename -o rename.csv gfa -o no_self_match_renamed.gfa > no_self_match_renamed.paf
minimap2 long_read.fasta long_read.fasta | fpa drop -l 500 index -t query -f match_upper_500.paf.idx query > match_upper_500.paf
minimap2 long_read.fasta long_read.fasta | fpa -o match_upper_500.paf.bz2 -z bzip2 drop -l 500 index -f match_upper_500.paf.idx -t target 


Only the mapping passed the filters are analyse by generators


The rename subcommand replaces the name of the read with another one.

If you use -i option the file will be read as a two-column csv, the first column is the original name and the second corresponds to the new name:

original name1, new name1
original name2, new name2

If the name of the read does not exist in the file it will not be replaced.

If you use -o, the names will automatically be replaced by a number a file like above example will be created.


fpa can build an index of offset of the records in the file where a reads appears.

The index file looks like this:

read_id1, start_of_range_1:end_of_range_1; start_of_range_2:end_of_range_2;…
read_id2, start_of_range_1:end_of_range_1; start_of_range_2:end_of_range_2;

fpa can index read only when it's query (first read in record) or target (second read in record) or both of them.


fpa can generate an overlap graph with overlap pass filters


  • Rust >= 1.32
  • libgz
  • libbzip2
  • liblzma


With cargo

If you have a rust environment setup you can run :

cargo install fpa_lr

With conda

fpa is avaible in bioconda channel

if bioconda channel is setup you can run :

conda install fpa

From source

git clone https://github.com/natir/fpa.git
cd fpa
git checkout v0.5

cargo build
cargo test
cargo install


~135K SLoC