Size of Flanking region genes during AUGUSTUS training

[ConcettaDe4]

Hi! I am using AUGUSTUS to predict genes in a new species. I am following the current protocol of AUGUSTUS and I extracted the flanking regions of the genes of my training and test set. I was wondering how the flanking regions affect the training and the prediction accuracy. Is it better to give a short region, as 100 nt or a larger region as 1000 nt?

Thank you for your help!

Concetta