SORAD SORAD: A systems biology approach to predict and modulate dynamic signaling pathway response from phosphoproteome time-course measurements usage: data = Sorad(data) input (the fields of the data structure): data a cell matrix (number of proteins by number of conditions), where each element has a row vector (number of time points) containing the protein and condition specific training data protNames a cell vector containing the names of the proteins conditionNames a cell vector containing the names of the training conditions viz a boolean variable controlling the visualization infer a boolean variable controlling whether to calculate the marginal likelihoods (true) or predict the dynamics of the system given the topology (false) IF infer == true maxRegulators a variable controlling the maximum number of regulatory proteins (<= number of proteins) IF infer == false networkMatrix a binary matrix containing the connectivity matrix of the signaling pathway, prot_j -> prot_i := (i,j) testData a cell matrix (number of proteins by number of test conditions), where each element has a row vector (number of time points) containing the protein and condition specific test data conditionNamesTest a cell vector containing names of the test conditions output (the fields of the data structure): X a cell vector (number of proteins) containing the fitted x functions (number of time points by number of conditions) f a cell vector (number of proteins) containing the fitted f functions (number of time points by number of conditions) param a cell vector (number of proteins) containing the protein specific alpha and lambda parameters IF infer == true subPowerset a cell vector containing the model definitions, in other words, the regulatory proteins of the different models margLikelihoods a matrix (number of models by number of proteins) containing the calculated marginal likelihoods of the models (the order is the same as in subPowerset) IF infer == false hyperparameters a cell vector (number of proteins) containing the logarithms of the protein and model specific sigma, l and sigma_n^2 hyperparameters trajectories a cell matrix (number of test conditions) in which each element has a matrix (number of time points by number of proteins) containing the predictions notes: - please see example.m for an example of the use - it might be a good idea to scale the data in such a way that it is between 0 and 1 - current implementation requires that all the measurement sets share the same time points - current implementation uses the time points in the training data also in the prediction - implementation of the intervention estimation procedure for a general setting is not straightforward so it is not included in this routine, however, upon a request the author will provide an example, which can be modified based on the given requirements The GPML Matlab code is written by Carl Edward Rasmussen. The simulated annealing implementation is written by Joachim Vandekerckhove. Author: Tarmo Äijö Updated: 07/21/2013