Confidence with I-TASSER prediction

Dear all,

I’m hoping others, with significantly more experience than I, will be able to suggest whether the combined TM-score and C-score from an I-TASSER prediction are “good enough” for further study.

I have a transmembrane protein, approximately 1800 residues in length (it’s a big protein). Sadly, homology modelling efforts have failed given the lack of similar crystal structures. As I am only interested in the extracellular binding domain, which is approximately 500 residues in length I removed everything else and once again tried homology modelling. There was no difference, no surprise there. So I’ve just had results back from I-TASSER as my last effort:

Model 1
C-score = -3.09
Estimated TM-score = +0.37+/-0.12
Estimated RMSD = 15.2 +/- 3.5 Angstrom

Model 2
C-score = -1.93 (but breaks the protein into two separate domains)

Model 3
C-score = -1.65 (a single protein)

The TM-score in model 1 is border line OK (almost 0.5, as per the documentation), however, I am very concerned with the low confidence C-score and large RMSD.

Am I looking at a very poor structure?