I have to accomplish an homology modeling task in which the reference template does not cover very large loops of the query sequence (20-30 residues). I have not found any reliable alternative template for that disordered regions. Fortunately, they are located far away from the site of interest for my project. Nevertheless, the softwares (ie MODELLER, GROMACS) and the analyses I have to carry on need an entire ungapped protein model. I thought about substituting the large missing loops with short ex-novo peptidic patches (such as a poly-ala sequence). In such way I will reduce the background noise generated by sub-optimal flexible regions created ab-initio. I am looking for any suggestion that could corroborate such an idea (ie a validated method and/or literature references)