Deconvolution accuracy, by AUC and correlation, in the presence of various kinds of experimental noise. a) Accuracy when some elements of the sort matrix are incorrect. b) Accuracy with the sort matrix perturbed by removing some cells from all measurements. (In a) and b), the x-axis represents the total number of entries in the sort matrix which were perturbed.) c) Accuracy when deconvolving with measurements perturbed by random noise (three different averages for each noise level are shown).