I have been using braker3 to annotate a novel genome. It is fast and so far has been pretty accurate except one problem. 5% of the time, it will fail to predict a gene with sufficient hints file evidence from RNA-seq and protein homology. Before moving to other tools, I’d like to see whether I have explored the full potential of braker.
The first option that came to my mind is --alternatives_from_evidence based on this discussion. However it turns out that this option is true by default.