Don't know what exactly you want to do: in the past we used %of most frequent occurring nt at that position. For amino acids (I think it will work for nt as well) you can use a variability plot: the number of different nt occurring at that position divided by the frequency of occurrence of the most occurring nt. 

If you want to compare the diversity among samples, you should  rarefaction, that is to take down the sample size to the lowest in your dataset. You will loose power but avoid bias. Several other analyses can be done without rarefction. However, there is still a lot of contraversy over these matters.

I agree with Serge, this totally depends on what you want to do

I agree with Jan. You should reduce your number of sequences to the lowest one, occuring in your samples.  This should be done such that the reduction is done randomly. I have till now worked with MOTHUR to do that using the command "sub.sample(fasta=sample1.fasta, size=x)", where x is the lowest number of sequences in the set of samples.

I think the best option is to import your results in MetagenomeSeq package and normalize. MetagenomeSeq's normalization method can take care of sequencing based unevenness... you can export your normalized results in a matrix and can use it further for any of analysis..

As others have said, it depends on what you want to do. In many cases, doing nothing is an option as it appears that rarefying has a negative impact on the accuracy of some indices (See link). If you are just visaualising what is there, just express everything as a percentage of sequences in that sample.

Article Waste Not, Want Not: Why Rarefying Microbiome Data Is Inadmissible