You certainly can go ahead: whether you should is another matter entirely.  If you have reason to believe that the catalytic core is largely independent of the rest of the protein, function wise, then that homology model should be reasonable.  That being said, homology modeling is a notoriously difficult practice in general, and may not provide substantially better docking results than simply using the same domain from a homologous protein.

If there are any other proteins in the literature where inhibitors for your protein also inhibit that other protein, and the other protein also has a crystal structure, I'd just do your docking on that (assuming that's what you plan for a virtual screen, anyways).  It skips the homology modeling step, which unless you're very comfortable with the technique is unlikely to give you anything particularly accurate anyways.

It depends actually on the quality of your model, the sequence identity between the template and that catalytic domain usually for a good quality we accept not less than 40%, but for further screening we should expect not less than 70% of identity, then you have to check torsion angles, internal energies and local geometry of the model to avoid steric clashes which if found you would better do a refinment process. you would need to superimpose the model with the template as well. so as the principal of virtual screening is the structure based drug discovery, so this structure shouldn't be virtual. 

i got the homology model of the catalytic domain based on 39% identity and 100% query coverage. the model was checked in prosa and procheck which gave the positive results. furthur the model was relaxed for 5 ns short md.in that scenario can i go ahead?

while the identity is not that potent, but the quality estimation indicates a quite reliable model, if you would further plan any binding assay validation that would be too recommended to consider.

Good Luck!

My focus is htvs primarily.with the above suggestions is it go ahead with it , later studies could be binding energy estimation...

well, it's fine based on your plan, but let's add more refinement to your structure by validating using one of more servers like :

http://molprobity.biochem.duke.edu/

http://services.mbi.ucla.edu/Verify_3D/

http://sysbio.rnet.missouri.edu/3Drefine/

http://www.reading.ac.uk/bioinf/ReFOLD/ReFOLD_form.html

http://zhanglab.ccmb.med.umich.edu/ModRefiner/

http://zhanglab.ccmb.med.umich.edu/FG-MD/

it would be also recommended to perform monte carlo sampling for refinment 

http://inka.mssm.edu/~mezei/mmc/

Thank you Mohammed for your detailed comments and resources !!!