Table 1. Summary of the Theoretical Ions and the Features Computed from PSM^a.

	fragment site	productions
1	backbone	b, b – water, b – ammonia
		a, a– water, a – ammonia
		y, y – water, y – ammonia
2	backbone closest to C-terminus	b + water, a + water
3	between lysine side	precursor, precursor – water, precursor – ammonia
	chain and the linker	precursor + linker, precursor + linker-water precursor + linker – ammonia, precursor + linker – double water
	features
1	number of cleavage sites supported by a matched prefix ion (i.e., b and a)
2	number of cleavage sites supported by a matched suffix ion (i.e., y)
3	length of longest consecutive cleavage sites supported by a matched prefix ion
4	length of longest consecutive cleavage sites supported by a matched suffix ion
5	binary value indicating if a precursor ion (e.g., precursor + linker) matched an experimental peak
6	percentage of experimental peaks that can be assigned
7	percentage of total experimental peak intensities that can be assigned
8	peptide length

^aGiven PSM (α, β, S), the masses of theoretical ions are first computed by assuming a single cleavage at each peptide bond for α and β, which leads to the regular b, y, and a ions. In addition, we consider the ions resulting from the dissociation of the bond between the side chain of the cross-linked residue and the linker (e.g., the bond between lysine's amino group and DEST). The matching between the theoretical ions and the peaks in S are represented as feature vectors for α and β separately, giving rive to x_α and x_β (each having a length of eight).