The easiest form of input is simply to make a plain text file with a set of FASTA data like those we have been using previously. You can get Genbank or GenPept to display the DNA/protein sequences in FASTA format and just c+P them into a txt file. You can even get a set of FASTA files throught the Homologene database.
If people are really interested in bioinformatics then I would definitely recommend downloading the Phylip suite of programs. It is fairly fiddly and technical to use all of the programs to do exactly what you want but the analyses produced are considerably more powerful and sophisticated than the sort of things we have been doing with Clustal in terms of phylogenetics.
Another nice new program has just been released. V2 of Jalview is out now and while it is not the best program for doing your actual alignments it allows you to visualise pre-existing alignments quite nicely and does some quite cool things with the trees, although it won't let me specify an outgroup for some reason. It also lets you run sequences through ClustalW and will display your data as a principal component analysis plot.
I would recommend the PAUP* package. It is not free ($85 when I purchased mine 2 years ago), but it is easy to use and offers a wide variety of programs and parameters with which to analyze data (nucleotide, protein, or morphological (coded)).
Ned asked how such data is used to generate trees. In short, parsimony and likelihood algorithms analyze the data for patterns of nucleotide substitution. Indeed, the degree of similarity is ignored in such programs for identical nucleotide/amino acid sites are irrelevant. Distance methods do use 'similarity', but I do not use such methods much. Amino acid sequence data is not used in phylogentic analyses nearly as much as it used to be for a couple of reasons - DNA data can provide at least 3 times the phylogenetic informton that amino acid sequence data can (hypothetically, providing we are only using protein coding sequence). Of course, non-coding DNA is usually much more phylogenetically informative in that it can accumulate more substitutional change than can conserved sequence (such as protein coding sequence).
Making up the input files for these programs can be tedious and frustrating. PAUP, for example, will produce an error message if you have misplaced punctuation (certain symbols are used in tghe input files - e.g., a ";" is used to denote the end of a data block) but it will not tell you where the missing symbol is (at least the earlier versions did not - I think the new one does). Someone had mentioned making plain text files - that usually works. I am pretty lazy, so when I am making a new input file, I usually just use an old one that I know works and cut and paste the new data into it.
As for the supposed anomalous trees using cytochrome C and B, immediately the use of amino acid data tells me not to put much stock in it, plus the fact that as has been mentioned, they represent only two small loci (mitochondrial loci at that, which are known to in genral mutate faster than nuclear genes).
I agree with WK re: the use of Clustal for analysis.
For one thing, we are assuming that the alignment Clustal produced is optimal or at least very good. In my experience with Clustal, it produces good starting alignments that then need to be re-done by eye. Of course, I am used to doing alignments with 20 to 45 species each with up to 12 thousand nucleotides. It may work perfectly for 30 to 100 amino acids, but when you start tossing in big indels and such in huge nucleotide files, it starts spitting out weird results. I recall once putting in just 2 sequences and the result it gave me was one entire sequence in a row, followed by the second entire sequence - no alignment at all. Whoever wrote it is right - there are all sorts of parameters you can fiddle with that can help avoid problems like that, even so, with big files, I have found alignment programs of several types only good for getting a starting point.
And, if you are suign a questionable alignment, then one should expect any results from any analyses to be odd.
S. Wolfram, in the book MATHEMATICA, makes the claim that seems to me somewhat correct, that there will never be a Maxwell of Kant's grass biology able to "unscramble"(Wolfram's scientific term) a volume of gas but I can not get out of my mind the possiblity that RNA could be demon unscrambling"" 1-D symmetry projections to effects on molecular motion in the enviornment of DNA and proteins by means of the analyticity of the other kind of 1-D symmetry under rules from the 2nd law of thermodynamics quantum mechanically.
If this kind of reverse engineering was congizable sequence comparisions should indicate configurations of nanotechnology more desirable than those guessed at by starting from random walks.
It would then be interesting to develop means of relating DNA and protein comparisons to a similar distance metric. Perhaps that has already been done. I am not an expert in bioinformatics.