Your product will have millions of copies of your DNA of interest, but each copy represents one allele from one cell of the starting material (eg. if you had 100 cells there would be 200 possible alleles). If you sequence it directly you'll get a mixture T's and C's in your sequencing results at each CpG site, since some alleles are methylated and some are unmethylated, and you won't be able to differentiated between a methylated or unmethylated state at any given site (unless it's 100% unmethylated or 100% methylated). When you clone the PCR product, a single piece of DNA gets cloned into each bacteria cell so each colony will represent a single allele from your starting material. You then have to PCR amplify from each colony to get enough DNA for sequencing, but in this "colony PCR", you will end up with millions of copies of that one allele so it will be "pure" for either C or T at each CpG. If you colony PCR amplify and sequence 20 colonies then you're generally looking at 20 representative alleles likely from 20 different cells, and you can see what proportion are methylated and what proportion are unmethylated. Note that sometimes you can get the same allele in more than one colony since your original PCR step generated many copies of each allele. If you sequence 20 colonies it might actually represent 18 alleles plus two PCR duplicates. To differentiated between alleles you should look for variation in CpG methylation and conversion at each C, including non-CpG C's. It is unlikley that two different alleles will have the exact same patterns for both of these things.
I hope I explained that ok, it's kinda confusing. I don't know any specific reading material I just learned bisulfite sequencing from others in my lab and the lab next to us. Feel free to ask for any clarifications.
Edited by biznatch, 28 May 2012 - 04:57 PM.