Fig. 3.

Approximation of joint probability. Because it is computationally expensive to compute the exact joint probability for a mixture spectrum (scales exponentially with the number of peptides), we approximates it using the product of single-peptide and conditional probabilities, which can be computed in linear time. As shown in A, for most cases the product probability accurately approximate the joint probability (most points cluster tightly around the main diagonal). For correct-matches, there are some cases falling below the main diagonal showing that the approximated probability is sometimes lower than the true joint probability for correct matches. However, as shown in the Precision/Recall curve in B, for the practical purpose of distinguishing correct-matches and incorrect-matches, the two distributions remain very well separated using either the exact joint probability or its approximation.