Preprocessing

The csRNA-Seq data undergo minimal preprocessing: the read alignments against the respective genome assemblies are filtered to retain only uniquely mapped reads, and 5' ends of the reads are used to assemble the genome-wide read count profiles. The rest of the preprocessing is related to setting up the benchmarking.

1. First, for each chromosome or contig, we select the genomic regions that do not overlap with the repeats. Regions shorter than are discarded.
2. Second, we assign the individual chromosomes or contigs to training or test data to maintain the balance in GC composition (considering only non-repeat regions).
3. Third, we subset nearly half of the test data for scoring during the online leaderboard stage (the complete data will be used for the final evaluation).