sliding_window build bin_paths.txt --threads 8 --output index.out --size 80m

sliding_window search --index index.out --query example_data/64/reads/all.fastq --pattern 50 --output matches.out --overlap 10

The aim of this repository is to develop an IBF based prefilter for metagenomics read mapping. The IBF is created from the (k,k)-minimiser content of the reference database. The filter excludes parts of the reference datbase for each query read. Only reference sequences where an approximate local match for the query sequence was found are retained. A local match is defined as:

• length >= w
• errors <= e

where w is the window length and e is the allowed number of errors. Each read is divided into multiple possibly overlapping windows. The (k, k)-minimiser content of each window is then queried in the IBF.

git clone --recurse-submodules https://github.com/eaasna/sliding-window

cd sliding-window
mkdir -p build
cd build
cmake ..
make

export PATH=$(pwd)/bin:$PATH
raptor --version

A toy data set can be found here.

wget https://ftp.imp.fu-berlin.de/pub/seiler/raptor/example_data.tar.gz
tar xfz example_data.tar.gz

After extraction, the example_data will look like:

$ tree -L 2 example_data
example_data
├── 1024
│   ├── bins
│   └── reads
└── 64
    ├── bins
    └── reads

The bins folder contains a FASTA file for each bin and the reads directory contains a FASTQ file for each bin containing reads from the respective bin (with 2 errors). Additionally, mini.fastq (5 reads of all bins), all.fastq (concatenation of all FASTQ files) and all10.fastq (all.fastq repeated 10 times) are provided in the reads folder.

In the following, we will use the 64 data set. To build an index over all bins, we first prepare a file that contains one file path per line (a line corresponds to a bin) and use this file as input:

seq -f "example_data/64/bins/bin_%02g.fasta" 0 1 63 > bin_paths.txt
sliding_window build bin_paths.txt --threads 8 --output index.out --size 80m

You may be prompted to enable or disable automatic update notifications. For questions, please consult the SeqAn documentation.

Afterwards, we can search for all reads from bin 1:

sliding_window search --index index.out --query example_data/64/reads/mini.fastq --errors 2 --pattern 50 --output matches.out --overlap 10

Each line of the output consists of the read ID (in the toy example these are numbers) and the corresponding bins in which they were found:

0       0,
1       0,
2       0,
3       0,
4       0,
16384   1,
...
1015812 62,
1032192 63,
1032193 63,
1032194 63,
1032195 63,
1032196 63,

For a list of options, see the help pages:

sliding_window --help
sliding_window build --help
sliding_window search --help

The sliding window filter is based on Raptor. Raptor is being developed by Enrico Seiler, but also incorporates much work from other members of SeqAn.

In your academic works (also comparisons and pipelines) please cite:

• Seiler, E. et al. (2020). Raptor: A fast and space-efficient pre-filter for querying very large collections of nucleotide sequences. bioRxiv 2020.10.08.330985. doi: https://doi.org/10.1101/2020.10.08.330985

The subdirectory util contains applications and scripts related to the paper.

Raptor is open source software. However, certain conditions apply when you (re-)distribute and/or modify Raptor, please see the license.