Comment:
Pinned Results
This app simulates ion channel currents using hidden Markov models. If you load experimental data, it can estimate model parameters from single-molecule and ensemble recordings.
Data may be in QUB QDF or DWT formats, Axon/Molecular Devices ABF and ATF formats, AxoGraph AXGR and AXGX formats, Bruxton/TAC Acquire format, HEKA Patchmaster Pulse format, or in columns of text separated by comma or tab (not by spaces). e.g.:
Current,Voltage 0.01,-80 -0.012,-80 0.023,-80 0.008,-80 0.011,50 0.53,50 0.74,50 0.92,50
A blank line is interpreted as a segment break.
The Transition State (Committor) is the state in a linear model from which either terminal state becomes equally likely at the measuring duration tmeas
This is a prototype of QUB software running in the web browser. It simulates the output of a hidden Markov model, representing a physical system such as an ion channel, and recovers model parameters from experimental data. Our plan is to host a database of projects and associated data, and to accelerate computation using on-demand and cloud resources. Build #
HJCFIT likelihood function provided by DCPROGS/HJCFIT.
Idealize a record to find the optimal sequence of dwells (for this particular model), using Segmental K-Means (SKM).
Qin, F. 2004. Restoration of Single-Channel Currents Using the Segmental k-Means Method Based on Hidden Markov Modeling Biophys. J. 2004 86(3):1488-501
The committor is calculated (by optimization) from the A matrix as the position in the Markov chain in which the probability of reaching each absorbing end state is equal after a time T.
An N-state linear Markov chain has two terminal states and N-2 intermediate states. We number the states in the order they are connected, from terminal state 1 to terminal state N. The committor is the initial state from which the system is equally likely to enter either of the terminal states first. In the particular models we are considering, the terminal states have exit rates several orders of magnitude smaller than non-terminal states. When the process reaches a terminal state, it tends to stay for a long time (on average, the inverse of the exit rate). For these models we can solve an equivalent problem: pick a measurement time Tm < (1/min(terminal-exit-rate)), and find the initial state which gives rise to equal terminal occupancy probabilities at time Tm. For an initial state C, we form the entry occupancy vector P0 with P0[i] = {1 if i=C, 0 otherwise}. If C is not an integer, we assign entry occupancies proportionally: P0[ipart(C)] = 1-fpart(C); P0[ipart(C)+1] = fpart(C). We also define the measurement occupancy vector Pm = P0 * A, where A = e^(Q*Tm), and the fitness function f=(Pm[1] - Pm[N])^2 which is minimized when Pm[1] = Pm[N]. We minimize f numerically with respect to C to find the optimal committor.
| Name |
|---|
To see dose-response plots, add a stimulus to one or more rate constants, and set up an appropriate simulation. Checking I = f(V) adds implicit stimulus named "Voltage."
To make a rate constant stimulus-dependent, first click it, then look under Rate properties. For concentration-dependence (linear, pre-exponential), enter the name of the substance next to Ligand name. For voltage or pressure sensitivity, check the appropriate box and enter a nonzero number next to k1 or k2 respectively.
The Simulation menu, below, generates various stimulus waveforms. Dose-Response builds a stimulus ladder. Agonist-Antagonist adds a second ladder variable. Paired Pulse varies the duration between an initial conditioning pulse and subsequent test pulse. In Steady-State mode, all model variables are held constant.
This panel will show the peak and equilibrium current at each stimulus level. Only the selected data, shown in the hi-res (lower) panel, will be measured.
Use the Data menu (or the icons at left) to adjust signals, subtract baseline, etc. If your data were recorded with varying voltage, pressure, or ligand concentration, but that signal is not in the file, use the Stimulus Builder to re-create the waveform.
To analyze, pick Macroscopic or Single-Molecule mode from the QUB menu.
To start analysis, pick Macroscopic or Single Molecule from the QUB menu.
Single molecule data visits a small number of discrete levels. We idealize it using SKM (Qin 2004), make duration histograms, and solve kinetic parameters with algorithms such as MIL (Qin et al 1996,1997).
Macroscopic data has the sum of dozens or thousands of single molecules. We solve the kinetics directly with algorithms such as MAC (Milescu 2005).
Chop Idl uses idealized data to select batches or bursts of events.
To use this function, first idealize the data using Single Molecule mode (in the QUB menu). Then come back here by choosing Data Entry from the QUB menu.
Switch to Single Molecule mode now?
States are represented as wells, connected by traversing an energy barrier.
More info here soon...
This page divides a dataset into clusters using the K-Means algorithm. "X-Means" refers to the question of which K is best—how many clusters? For each K, we run several randomized trials to find the optimal clustering, and compute the Sum Square Residual (SSR) and corrected Akaike Information Criterion (AICc). To see a particular K-clustering, click one of the SSR or AICc results.
To get started, use the File menu to load or enter data.