Algorithms
From QuB
| Prev: Overview | Outline | Next: Requirements and Assumptions |
Contents |
Simulation
Simulation generates microscopic and macroscopic currents from a markov model. Currents can be stimulated, with ligand- and voltage- dependent rates, and voltage-dependent conductance. Various noise and artifacts can be added for verisimilitude.
SKM
SKM (Segmental K-Means) detects opening and closing transitions in raw channel data. The User makes a simple model (generally using one state for each conductance class) with starting amplitude guesstimates entered by highlighting the salient portions of the curent record. SKM uses a hidden Markov algorithm to produce a list of idealized open and closed interval durations for the selected data. For each conductance class SKM estimates a mean amplitude, standard deviation, duration, and occupancy probability. In addition, the likelihood of the most likely state sequence is provided.
Because SKM incorporates a noise model into the detection algorithm, small-amplitude currents can often be detected without invoking severe low-pass filters.
SKM also provides the average amplitudes of events, an amplitude histogram, the probability of the channel being in any state in the model, the mean lifetimes, and the approximate log likelihood for both the model selected by the User and the input data.
Idl/Base
Idl/Base is SKM combined with a Kalman filter that can track significant baseline fluctuations.
MIL
MIL (Maximum Interval Likelihood) optimizes the rate constants (with error limits) of a User-defined kinetic model according to the interval durations detected by SKM (or some other program). The kinetic model will typically have more than one state for each conductance class. A correction is made for 'missed events', which is valid for both single-level and multi-channel data.
The parameters of the model may be constrained according to detailed balance, fixed rate constants, and proportionality of rate constants. Multiple files, derived from data obtained at different voltages or concentrations can be analysed by MIL simlutaneously to estimate the voltage or concentration dependence (with error estimates) of all rate constants.
Other output from MIL includes event duration histograms for each state, the predicted probability density function (PDF) for the event durations, and the log likelihood for the model.
MPL
MPL (Maximum Point Likelihood) maximizes the likelihood of the data at every data point, instead of every interval. The advantage of treating the data in this way, compared with interval methods, is that very busy data (data with residence times on the order of the sampling time) can be analyzed.
In this technique there are no missed events, but the penalty is that it takes longer than working with idealized data. MPL can actually solve the kinetics of data covered by noise. We recommend using it for estimating conductances, characterizing the noise of states, and for kinetic analysis of short pieces of busy data.
MIP
MIP (Maximum Idealized Point-likelihood) is much like MIL.
Mac
Mac estimates rate constants and other parameters for macroscopic data -- recordings with hundreds of channels.
Staircase
Staircase estimates rate constants for staircase processes, where the observed variable moves in one direction, with each step governed by the same Markov process.
Model Search
Model Search calculates the likelihoods of all possible models of a specified number of states and conductance classes. The program gives you an ordered list of the models and their likelihood. You can rerun models with random starting values to reduce the chance of local maxima.
Graph theoretic approaches have eliminated the isomorphic models (different drawings of the same interconnection patterns). This problem explodes with the number of states and at the moment is limited to models of six states (eight states may have 2 x 106 possible connections).
While you can apply Model Search to a set of data files obtained across concentration or voltage, the program will not distribute the dependencies to particular rates. If the model is ligand dependent, then every rate is considered to be ligand dependent....a bit messy. We recommend using Model Search as a rough guide to possible state models for your data and then to explore the details more closely using MIL. (The core of Model Search is MIL).
Model Merge
Model merge builds a new model by combining two models. The rate constants are constrained to keep the two sub-models independent. You can relax constraints to model cooperative behavior, or "un-merge" solved rate constants back into the original models. (in the model's right-click menu)
| Prev: Overview | Outline | Next: Requirements and Assumptions |
