I want to create a full model of 1899 residue long receptor protein. The protein has only 19% identity with PDB proteins and having a middle transmembrane and C-terminal domain. To model it with I-Tasser, I need to break my query sequence. My question is - could you let me know about the publication of I-Tasser where there is some kind of evaluation of protein model quality with respect to the protein length? Based on performance of I-Tasser, for the better modeling how long I should break my query sequence?