Try using i-Tasser. 

http://zhanglab.ccmb.med.umich.edu/I-TASSER/

You are making a prediction from another prediction, which is tenuous.  I would suggest crystallizing the viral capsomer and determining its 3D structure by x-ray crystallography.  You can then use that experimental (not predicted) 3D structure to predict the  higher-order structure.

If you can figure out which template was used to make your model, you can align your model with that one if it is a full capsid.

Dear Dr.Reyes

Thank you for suggestion, but I have no access to x-ray cristallography instruments. However, as @Joosten mentioned, I have predicted capsomer 3D structure with 87% identity using a known homolog as template. Then I used UCSF Chimera to make a high-ordered structure. Overall, I figured the predicted structure is only on of the n-homomer making capsid protein.

You can also try the ClusPro server (https://cluspro.org)