Genome Annotation with Maker - How Much Evidence Is Enough? How Much Evidence Is Too Much?

De novo gene prediction programs are much more accurate when provided with evidence to help guide gene predictions.  The MAKER gene annotation pipeline simplifies the process of aligning transcripts and proteins to genomic sequences, generating hints for de novo prediction programs, running de novo prediction programs, and picking gene models with the highest concordance to existing evidence.  It is now relatively cheap and easy to produce transcript data to use as evidence for gene annotation for new genomes.  However, researchers working with newly sequenced genomes from species with few or no transcript sequence resources may have to generate those sequences before genome annotation can take place.  Despite the fact that sequencing costs are continuing to fall, the effort to generate a large number of RNA-seq libraries and assembly them can still be substantial.  Knowing how much transcript evidence is actually necessary to generate a high quality genome annotation is therefore prudent.  On the other hand, given that RNA-seq data is relatively cheap to generate, there is a question as to whether too much transcript evidence will reduce the quality of a new genome annotation.  New evidence sets always introduce spurious sequence alignments that can potentially confound gene prediction software.  We have explored these issues of evidence quantity and annotation quality using the MAKER genome annotation pipeline, and we will present a set of guidelines to help researchers make informed choices for the annotation of new genomes.