Lessons learned from manual curation of thousands of gene models in the nematode Pristionchus pacificus

Continuous developments in sequencing technologies have led to the generation of chromosome-scale genome assemblies across the whole tree of life, but our ability to annotate genomes has lacked behind. One major problem consists in the fact that typically not all genes are expressed at detectable levels at any given life stage or environment. Therefore, available transcriptome data needs to be complemented by gene prediction programs and protein homology evidence. However, how to optimally combine these different data types is not well understood. Here, we present a case study, where we community curated gene annotations of the Pristionchus pacificus strain RSC011. By incorporation of new Iso-seq and RNA-seq data and genome-wide screening, we identified and corrected more than 7,500 ([~]24%) gene models. While the improved gene annotation for the RSC011 strain will be useful for the P. pacificus community, our study reveals several gene annotation problems that may affect data from other species. Among these, we identified assembly errors, artificial transcript fusions resulting from overlapping genes and polycistronic RNAs, falsely called open reading frames, and error propagation based on homology data as frequent sources of gene annotation errors. Thus, our findings may be helpful in guiding future efforts to annotate genomes across different taxonomic groups.