Is parallelization possible in mcsqs?
and if yes, what is the optimum number of cpus?
It’s in the manual.
Just run multiple copies of mcsqs with a different number specified as the -ip option (to avoid file clashes).
The larger number of cpu the better. 
Thanks very much indeed.
Is it possible to use MPI in the job script (to be submitted to the queue) together with multiple cpus?
You can compile and use the code embpar (in atat/src).
You give it a filename containing all the commands to run in parallel (here mcsqs -ip=… etc), one per line.
Same question. How to use embpar or some other tool for running mcsqs on multiple nodes via MPI ?
cd atat/src
make embpar
cp embpar ~/bin/ #or wherever you put your binaries
create a file jobs.in containing e.g.
mcsqs -ip=0
mcsqs -ip=1
mcsqs -ip=2
mcsqs -ip=3
In your batch script:
mpirun -np 4 embpar jobs.in