The B-cell epitopes can be predicted by the ANTIGENIC program in EMBOSS (http://bioinfo.hku.hk/EMBOSS/). Potential Th epitopes can be predicted by the ProPred HLA-DR binding peptide prediction program (http://www.imtech.res.in/cgibin/hlapred)
For T cell epitope prediction, I recommend IEDB (http://www.immuneepitope.org/) but also caution that any prediction tool will have a high rate of false-positives.
May I ask the questions more specifically? If I have 10 strains genome of a bacteria from NCBI. Firstly, I want to analyze to find a list of genes that are homologous in these strains. Secondly, I wonder to screen size below to 30-50 kDa and annotate their functions. What tools could I do? Thanks.
as for your second question, you can used a Perl script to extract all predicted protein from your bacterial genome (GenBank). and then predicted the protein size in batch module using ExPASy
I they are just strains, maybe a Reciporcal Blast Best Hit is another option. You can use OrthoMCL too, but is a little hard to install. For annotation, hmmer coul be an option if you don't have so much proteins or Blast2Go.