Figure 4
Sampling approach: Representing spectral counts with ensembles of binary matrices. A summary of our sampling approach. First, each spectral count in the AP-MS data matrix is converted to a probability 1 - (1 - p)n, where n is the spectral count. Then, for each cell of the matrix, we sample an independent Bernoulli random variable according to its probability. We repeat this procedure independently for a desired number of trials, obtaining an ensemble of binary matrices representing the original quantitative AP-MS data. Each binary matrix is then used as input to a PPI inference method of choice that operates on binary data, and the results from each trial are aggregated to produce an ensemble score. Notably, the existing PPI inference method is directly applied to each binary matrix without modification.