As far as I can see you have a set of nonsynonymous recurrent mutations in a group of genes. The reasons why you should also include the amino acid number is explained by the colleagues here.
I wanted to little bit help about the comparison you wanted to make. You mentioned earlier that you wanted to compare them in means of amino acid type. One way I can think of this is to look at the distribution of both amino acids along your protein, let's say in windows of 10 residues. And then you can look at their log2 ratio near the site of your mutation. If the ratio is high (let's say 1.5/-1.5) then you can now look at the blosum62 (the matrix choice here can be changed) scores for those two amino acids giving you an idea (only an idea!) about their interchangibility. You can repeat this procedure for the other genes and take a look at your results. If the nearby log2 ratios are similar in all the genes harboring mutations involving those 2 amino acids than you have a hypothesis to buildup. But It is not clear to me whether these alone would be enough. I suggest you back up your work with additional data.
You can do the aforementioned work with a software we have recently published (http://i-pv.org/). I give an example below from FOXP2, ratio of arginines versus lysines. The green graph is the blosum62 agreement graph, the orange one is the log2 ratios. You can check the gif file below:
To generate these graphs for your protein, you will need perl and circos locally installed with all dependent libraries. Then extract the ipv.rar file to anywhere in your PC. If you are lost with the installation let me know, I will generate the html file for your gene of interest.
I hope this helps,
good luck with your research,