ITM Probe
ITM Probe
Corresponds to absorbing Markov chains. Contains only sinks as boundary and for each node in the network evaluates the likelihood for a random walk starting at that node to terminate at each sink while avoiding all other sinks (hence the sinks can be thought of as competing for the flow). The total likelihood at each node is the sum of the individual likelihoods for all sinks. Due to dissipation, the total likelihood at each node can be much less than 1, especially for the nodes far from any of the sinks.
Contains both sources and sinks and combines emitting and absorbing models. It reports the total expected number of visits to any node in the network from a random walk originating at each source and terminating at any of the sinks (i.e. not dissipating).
A set of parameters that defines the environment for a random walk in an interaction graph. It includes boundary nodes (sources and/or sinks), rate of dissipation and the set of excluded nodes. ITM Probe evaluates the context to describe the information flow it represents.
Also known as termination/stopping probability. A parameter between 0 and 1 describing the proportion of random walks that leave the network (evaporate) at each step. The higher the dissipation rate, the more likely it is for the random walk to terminate close to its origin and thus have only local effects. In the context of the channel model, the dissipation rate controls the likelihood for the random walk to visit the nodes away from the shortest paths from sources to sinks – dissipation close to 1 means that only the nodes on shortest paths will be visited.
Contains sources on the boundary. For each node in the network evaluates the average number of time the node is visited by a random walk originating at each source. Random walks terminate when reaching any source or dissipating.
A protein node that is excluded from consideration during a run of ITM Probe.
In some cases, proteins with a large number of non-specific interaction partners might overtake the true signaling proteins in the information flow modeling. Therefore, ITM Probe allows users to specify nodes to exclude from the network. For the yeast network the nodes excluded by default include cytoskeleton proteins, histones and chaperones, since they may provide undesirable shortcuts.
Note that excluded nodes are treated as terminating points for random walks: the edges leading into them are not deleted but any random walk entering and excluded node evaporates instead.
A (weighted, directed) graph whose nodes (vertexes) represent agents and edges (links) are interactions. Thus, two agents are linked if they interact in some way. The weight of a link corresponds to the strength of the corresponding interaction.
A concept used in the context of the emitting and the channel model to describe the measure of overlap between visits of each node by random walks originating at different sources. Large interference implies large overlap between flows from different sources while small interference means little overlap.
Information Transduction Module. The set of most significant nodes (with respect to the number of visits, in the case of the emitting and channel models, or the absorption probability, in the case of the absorbing model) resulting from the query context of ITM Probe.
Used for the emitting and the channel model. Gives approximate number of nodes that have a significant (that is, much larger than ordinary) total number of visits.
An interaction network where nodes correspond to cellular proteins. The edges may represent a variety of interactions, for example: physical (protein A binds protein B), metabolic (A and B catalyze reactions involving the same chemical), genetic (A and B are expressed together) or biochemical (A post-translationally modifies B). It is also possible to include more than one type of interaction in the network, depending on the problem being modelled.
A mathematical concept involving an entity that moves about a given space in a random fashion. In the context of graphs, a random walk describe a process where a 'walker' moves from one vertex into another with a probability proportional to the weight of the edge connecting them. This process is equivalent to a Markov chain on the vertex set. A random walk starts at a node in a network and moves about visiting different nodes until it terminates. It can terminate either at a boundary point (source or sink) or by leaving the network due to dissipation.
A destination of a random walk.
A point of origin of a random walk.
see Dissipation rate.
see Absorbing model.
Used in the context of the emitting and the channel model to denote the total number of visits from all sources.