rob-p / split2samples

BD rhapsody sample splitting preprocessing

splitp

Streaming SPLiT-seq read pre-processing in Rust. Currently, this is just a Rust implementation of the Perl pre-processing script by @jeremymsimon here. In the future, capabilities may expand / become more general.

The raison d'etre of the program is simply to perform equivalent processing to the custom Perl script above, but much faster. To that end, here's a small benchmark on a set of 10,000,000 SPLiT-seq v2 reads. This is on a 2016 MacBook Pro 2.9 GHz Quad-Core Intel Core i7, 16 GB 2133 MHz LPDDR3 (but both programs are using a single thread).

program	runtime
Perl script	2m 48.6s
splitp	5.7s
splitp (pipe to /dev/null)	3.7s

Usage

The splitp program takes several arguments. The usage can be printed from the command line using splitp -h.

USAGE:
    splitp [OPTIONS] --read-file <READ_FILE> --bc-map <BC_MAP> --start <START> --end <END>

OPTIONS:
    -b, --bc-map <BC_MAP>          the map of oligo-dT to random hexamers
    -e, --end <END>                end position of the random barcode
    -h, --help                     Print help information
    -o, --one-hamming              consider 1-hamming distance neighbors of random hexamers
    -r, --read-file <READ_FILE>    the input R2 file
    -s, --start <START>            start position of the random barcode
    -V, --version                  Print version information

Please take note that splitp writes the output (processed reads) to stdout, so that the output can be directly piped to an input stream of another program (e.g. directly to alevin-fry via process substitution). This also means, if you want to store the processed reads on disk, you can pipe the results directly to gzip to compress them e.g.:

splitp -r reads.fq -b oligo_hex_bc_mapping.txt -s 87 -e 94 -o | gzip > reads.fq.gz

Notes

The input oligo-dT to random-mer mapping must be provided in a two-column tab-separated file. Further, the first row of the file must be a comment starting with the # character.

Limitations / differences

• Currently, random hexamers can only be searched at a Hamming distance of 0 or 1.
• If the barcode being considered for replacement (BC1) has N characters in it, they are all replaced with A before lookup in the table of random hexamer hamming neighbors. If there is more than one N, this may result in slightly different behavior than the Perl script.