File formats & submission templates

The training data are provided on Zenodo on the whole-chromosome or whole-contig level as a set of per-chromosome/per-contig bed files; the files contain the data for + and - strands, see the detailed description in a dedicated section.

The submission files must be plain text files with one real value (prediction score) per line. Prediction scores should be non-negative real values within the [0..1] range with up to 5 digits after the decimal point (i.e., 0.xxxxx). The submission file templates accompanied by the list of genomic intervals of interest (i.e., not overlapping repeats and located on the test-set chromosomes) are provided with the challenge data. For convenience, we also provide an analogous subset of intervals in the training data, so the users can account for this setup during model training or local validation.

Each line in the submission file corresponds to a particular genomic coordinate (position) and strand. A single submission file should be generated for each species, and the submission files should be individually gz-packed before submission (e.g., gzip my_submission_oyster.txt).

For more details, please refer to the detailed description of the submission file formats.