Thank you for your interest in my work. Unfortunately there is no accurate method for predicting Linear B cell epitopes currently available via the internet. Protein sequence profiling methods using Hydrophilicity, Flexibility, and Accessibility etc... do not work well. However if you choose to use one of these methods they can be found here: http://www.imtech.res.in/raghava/bcepred/
If your intention is to successfully predict only 1 or 2 epitopes form your protein I would recommend trying to find protruding loop or turn structures located on the surface of the protein. If there is a known 3D structure of your protein in the PDB database this should be quite easy. Programs such as DSSP or STRIDE can analyze your 3D protein structure and tell you were these regions are and how accessible they are. Available here: http://bioweb.pasteur.fr/seqanal/interfaces/dssp-simple.html http://bioweb.pasteur.fr/seqanal/interfaces/stride.html
If you do not have a known structure for your protein I would use a secondary structure prediction program such as JPRED or PREDATOR to predict the location of loop and turn structures within your protein
Once you have found a region of your protein that you think is likely to contain a B cell epitope I would also recommend synthesising a large peptide of approximately 15 amino acids to cover it as the local secondary structure maybe important in enabling an antibody to cross-react.
Finally, I do not know a great deal about recombinant protein technology but I would imagine that so long as a predicted B cell epitope is still accessible and is not masked from being fussed to another protein (i.e. The predicted epitope is not facing the join to the other protein) the fussing or your protein should not have a significant effect.