Telomere repeat prediction from k-mer counts

Predict telomere repeats using k-mer counts from unassembled read libraries.

This assumes that telomeres are the most abundant direct tandem repeat sequences in a given genome library. The output is a list of direct tandem repeat unit sequences, sorted by frequency.

k-mers are first counted with jellyfish count before being k-mer sequences and counts are dumped to Fasta format. Those above the minimum coverage are then retained, and used to search for the most abundant tandem direct repeats. An example bash script for this task is given in do_jellyfish.sh. These scripts have been tested with jellyfish v2.2.10. Dependencies listed in Conda environment file env.yml.

NB:

• k-mer length used for k-mer counting should be prime and longer than the maximum expected telomere repeat unit length.
• Minimum coverage cutoff supplied to discard_singletons.py should be slightly above the expected average coverage of the genome, to retain only repeat elements.

The JSON file containing counts of high-frequency k-mers is then supplied to the script find_repeats_from_kmers.py along with the k-mer length used for jellyfish count. Example:

./find_repeats_from_kmers.py \
  --maxzeroes 1 \
  -k 19 \
  --counts library.k19.cov100.json \
  --output library.k19.repeats