Modeling:Idealize

From QuB

Jump to: navigation, search
Prev: Actions:Modeling Outline Next: Modeling:Amps


Contents

Idealization

The Idealize (Idl) button describes sampled data as a sequence of events. This idealization is overlaid on the Data display, and is available to algorithms such as MIL which operate on idealized data.

Idealization also generates measurements such as the mean amplitude and duration of closed and open events, along with all-points amplitude histograms.

To idealize, we decide which class each data point belongs to, then join adjacent points of the same class into dwells (aka events or sojourns). Two methods of classification are provided:

Half-Amplitude
half-amplitude threshold detection
a data point is assigned to the nearest class, using only distance from the mean.
with two classes (closed and open) this is like drawing a line midway between the two means, and calling everything below the threshold closed, everything above open.
typically applied to filtered data
SKM
segmental k-means
Qin, F. Restoration of Single-Channel Currents Using the Segmental k-Means Method Based on Hidden Markov Modeling Biophys. J. 2004 86: March 2004
SKM repeatedly iterates two steps:
  • Viterbi detection of the most likely state sequence
    • For every time t, SKM computes the likelihood that the data came from each state, and the most likely previous state. The idealization is found by taking the most likely final state and following the chain of m.l. previous states.
  • re-estimation of rate constants and class amp/std using the MLSS
fails on over-filtered data

Both methods require you to prepare a Model and initialize (Grab) the class amplitudes. For details and examples, see

Related Actions

Properties

Data channel index which A/D channel contains the data to idealize, typically 0
Data source
Pre-process data none: use raw data
as displayed: use the same baseline correction and filtering as the Data display
as such: apply the specified filter and baseline
Thread count number of segments/selections to idealize at once. This will only have a significant effect if your Data Source has more than one segment or selection, and your computer has more than one processor.
Clear existing idealization completely remove any idealization, even in parts of the file that are not part of the Data Source. Otherwise, idealization outside the Data Source is kept.
Hist bin count number of bins for amplitude histograms, displayed after idealization in the Results window
Half amplitude selects the Half-amplitude (threshold) method described above
Compute stats (if Half amplitude is selected) generate all the measurements and histograms. Otherwise, Half-amplitude generates idealization only (which is much faster).
SKM selects the Segmental K-Means method described above
LL conv stop iteration if LL improves by less than this much
Max iter stop after at most this many iterations
Drop first/last event omit the first and last dwell from the idealization. Useful when idealizing selections that begin and end with partial events (such as a piece of a long closed event)
If class omit the first/last event only if they are in a certain class, typically class 1 (black/closed)
Re-estimate after finding the event sequence, re-estimate model paramters and repeat until max iterations. Otherwise, yield the first event sequence, using initial model parameters, and don't iterate.
Fix kinetics when/if re-estimating, don't re-estimate state transition probabilities.
Class (the following "fix" settings apply to this class. #1 is black)
Fix Amp when/if re-estimating, don't re-estimate the mean amplitude of this class
Fix Std when/if re-estimating, don't re-estimate the standard deviation of this class
Apply dead time to statistics compute statistics as though the dead time (main window upper-right) had been applied (by joining events shorter than the dead time with the previous event). This way stats such as "lifetime" will be more in agreement with MIL.
Apply dead time to idealization modify the idealization by joining events shorter than the dead time with the previous event. This is non-reversible. It's useful to see what MIL sees when it internally applies a dead time, but in general leave this un-checked so you can change the dead time later for MIL.

Results

In the textual Report window: the measurements of each segment/selection

In the Results window:

Summary:

amp the (re-estimated) mean amplitude of each class
sd the (re-estimated) standard deviation of each class

Segments (and Select, Criteria): all the measurements of each segment/selection

Models: the model used for idealization

Histograms: An amplitude histogram for each segment/selection, overlaid with the Gaussian distribution (amp, std) of each class. You can quickly identify segments that were mis-idealized by looking for histograms whose distribution curves don't match the histogram bars.


Prev: Actions:Modeling Outline Next: Modeling:Amps
Personal tools