RENOMAD

Implementation of NOMAD (https://doi.org/10.1101/2022.06.24.497555)

Binary file for Linux is in the bin/ sub-directory of this repository.

K-mer counting with KMC3

First run KMC3 for each sample, here is an example which assumes ERR5296631_1.fastq is present in the current directory.

SRA=ERR5296631
kmc \
  -k54 \
  -m32 \
  -b \
  ${SRA}_1.fastq \
  kmc_out/${SRA}.kmc \
  kmc_tmp/

The -k option sets the k-mer length from KMC's perspective. This is 2k from the NOMAD perspective, where k is the anchor length and the target length. Renomad splits the KMC k-mer into two equal-length NOMAD k-mers. Thus, there is no support for a gap between the anchor and target.

By default, KMC transforms k-mers into canonical form, i.e. reverse-complements the k-mer sequence if this is first in lexicographic order. This is disabled by the -b option, use this when reads are single-stranded e.g. RNA-seq.

Sort KMC3 output files into lexicographic order.

SRA=ERR5296631
kmc_tools transform $SRA.kmc sort $SRA.sorted.kmc

Run renomad

Create a text file named kmcfiles.txt with pathnames of the sorted KMC output files, one per line. Then run renomad as follows:

renomad \
  -joinp kmcfiles.txt \
  -maxp 0.05 \
  -tsv3out output.tsv

The maxp option is the maximum P value to report an anchor. Default is 0.05.

Output format is tab-separated text with three fields: (1) count, (2) anchor+target sequence, (3) sample_name. The sample name is extracted from the KMC filename by stripping the path name and extension.