We used Markov Chains in order to differentiate amongst two categories, a positive one and a negative one. With Markov chains we calculate the probability of a certain residue being in a certain position, given the preceding residue. Thus we create matrices with which we predict whether a sequence is positive or negative depending on the training datasets we had initially provided. For PredSL, we used several combinations of positive and negative sets, and therefore got a corresponding number of prediction scores, for each sequence. The combinations we used are:

Sequence Type Positive Negative
Plant chloroplast cytopasmic
mitochondrial cytoplasmic
chloroplast mitochondrial
chloroplast & mitochondrial cytoplasmic
chloroplast & mitochondrial secreted
chloroplast mitochondrial & cytoplasmic
Non-plant mitochondrial cytoplasmic
mitochondrial secreted
mitochondrial secreted & cytoplasmic