1. I think you should give more details about how you will model the
 probability of fragmenting at a particular location. It sounds as though
 Zhang's mobile proton model primarily uses the composition of the peptide
 (relative frequencies of basic residues and protons, and you will be
 somehow combining that with the effects of specific residues adjacent to
 the cleavage site. How will you combine these -- will you assume the
 effects are multiplicative?

 2. One big advantage of the Zhang model is that it is based on an
 underlying model of the chemical events. It is unclear to me to what extent
 you are going to try to interpret the probabilities you find in terms of
 the chemistry. You should clarify this (e.g. why, chemically, is the
 fragmentation more likely to occur adjacent to some residues than others).
 If you can provide an interpretation of your probabilities this would
 increase the scientific value of your model.
  A related point is that having some specific hypotheses about what
	influences the fragmenting might give you some insight into how to
	construct the probability model. For example, maybe you should be taking
	into account distances of the potential fragmentation site from basic
	residues or proton donors (does Zhang do this already?) Also, I presume a
	lot may be known about characteristics of various amino acids such as
	precise bond lengths, etc -- can you somehow work that into a hypothesis
	about which sites would be relatively more subject to nucleophilic attack?

	4. Can you give a reason why you think you can do better than Zhang's
	model, other than just that it does poorly in some instances? For example
	in Figure 1, what modifications would have improved the E1 prediction?

	p. 7: "is analogous to the advantages of .." : can you explain briefly
	what those advantages are?

	p. 9: Also, have you thought
	about whether there might be other similarity metrics which are better for
	your purposes than the one Zhang uses? (For example, metrics that are more
	directly related to the accuracy of your underlying probabilities).

	p.14: does one need to take into account "non-neutral losses" -- i.e.
	losses of small charged molecules like H+ or H30+.