EPIC data

Genomes and training data are available on Zenodo: doi: 10.5281/zenodo.22285753.

The EPIC data are organized identically for each genome in a folder named after the genome assembly:

${ASSEMBLY}/

Additionally, genome_metrics.json provides the sizes of the EPIC train/test subsets for each assembly.

1. The contents of the ${ASSEMBLY} folders are as follows.

The genome subfolder provides the genome reference sequences

genome/genome.fa.gz genome sequence, obtained from GenArk, gzipped fasta

genome/genome.fa.fai index for the genome assembly file

genome/assembly_report.txt genome assembly report (including contig identifiers), obtained from GenArk

2. The for_participants subfolder provides the key data to be used for model training.

csRNA-${CELL_TYPE_OR_TISSUE}-{r1,r2}-train.bed.gz and for_participants/csRNA-${CELL_TYPE_OR_TISSUE}-{r1,r2}-train-unfiltered.bed.gz

are gzipped bed files (one per replicate) with the per-position csRNA-Seq read counts (strands are separated and listed explicitly).

Designation of columns: contig, start, end (=start+1), id, readcount, strand.

The unfiltered file includes all positions from the train contigs, while the basic file includes only regions which do not overlap genomic repeats (mimicking those used in the scoring protocol for the test data).

EXTRAS

for_participants/whitelist.{train,test}.bed.gz contain the repeat-free train intervals + repeat-free test intervals (those will be used in the assessment).

for_participants/template.{train,test}.bed.gz contain bed-formatted templates for training data and for predictions

Note: the submissions should NOT be provided in bed files;
only gzipped txt files will be accepted.

csRNA-train.txt.gz contains pooled train csRNA-seq tracks.

rnaseq-${CELL_TYPE}-{r1,r2}-train.txt.gz and for_participants/rnaseq-train.txt.gz provide the control sRNA-seq data processed in the same way as csRNA-seq.

3. Two folders will be hidden from the competitors (except for the toy example Nematostella vectensis data) until the challenge ends.

The scoring folder contains the mask and bed files separating the intervals (test data subsets) used in the live online leaderboard and the final ranking.

The ground_truth folder contains the challenge solution (the csRNA-seq profiles for the test set of contigs):

csRNA-test.txt.gz test data "answers" (pooled replicates).

rnaseq-test.txt.gz matching sRNA-Seq data (pooled replicates). Not used in scoring but might be used in the post-challenge assessment.