Due to assembly limitations of this draft genome, some manually annotated gene models were split across scaffolds. Like complete gene models, each split part of a given gene has its own unique ID in the "original" data set, comprised of the gff file as well as fasta files for transcripts' cDNA (including any UTR: "cdna"), coding sequence nucleotides only ("cds"), and the protein translation ("pep").
In a second step, we manually stitched the split models together, creating one model per gene, and where the stitched model has its own unique ID. Since transcript stitching ignores scaffold coordinate positions, there is no gff, but there are all three flavors of fasta file (cdna, cds, pep) for the "merged" data set.
The remaining files are documentation in support of the stitching. The Excel spreadsheet tracks which original models were replaced by which stitched versions. In total, 121 original, split models were stitched into 55 new models, with up to 5 original models corresponding to one stitched gene model. The "split-removed" files are the set of complete models only (121 split models removed), and the "stitched" files are the 55 manually fixed new models only. "Merged" is then just the concatenation of those two data sets.