Z-Score and cutoff in protein sequence alignment

[hu xiche]

        I am using the BESTFIT program in the GCG package to align two
protein sequences.  I would like to evaluate the statistical significance
of such an alignment.  By using the -RAND option, I can get the averaged
quality of the "shuffled" random sequences and the standard deviation from
the output of BESTFIT.  But I don't know how to convert them into
"Z-score".  I will deeply appreciate your help in answering the following
questions: (1) How should the "Z-Score" be calculated? and (2) What is the
cutoff for statistically meaningful alignment in terms of "Z-Score"?
Any references to published journal articles or names of experts who will
be able to help will be very helpful.

        Thanks in advance,
        Xiche Hu