What is most likely to be used, Homology or ab-initio based method?

Question

[Neha shri]

Dear all,

Suppose if I have a database of protein sequences , and each sequence in it shares a sequence similarity of more than 50 % with a sequence for which crystal structure is already available in the PDB.

Is it likely that any sequence in the database would fold into a different structure despite having a homologue in PDB? Please help me understand which modelling method , Homology or Ab initio, one should go with and why? Would appreciate any help. Thanks in advance.