Single Channel Measurements

From QuB

Jump to: navigation, search
Prev: Tutorial:Simulation Outline Next: Tutorial:Single Channel Measurements


Before hidden Markov model (HMM) analysis, single ion channels were analyzed with measurements, histograms, and statistics. Several of these are available in QuB, including mean open amplitude, open probability, mean open time, and amplitude and duration histograms with curve-fitting.

Contents

Measurements

Amp i, Std i
the mean amplitude and standard deviation of data during events of conductance class i
occupancy
aka occupancy probability, Popen, Po, Pclosed
the percent of the record spent in one conductance class (color)
lifetime
the mean duration of events in one conductance class
first latency
time elapsed between the beginning of a data segment and the first channel opening.
nevent i
the total number of events of class i

Histograms are available at any time in the Histogram window (View -> Histogram). Right-click a histogram for options. Histograms are generated from the Data Source. To curve-fit any histogram, right-click it and choose Curve fitting.

Idealization

To get most of the measurements you must idealize your data. Idealization, also called event detection, describes the data as a sequence of open and closed intervals. QuB's idealization routines automatically calculate most of the relevant stats.

Which idealization routine should you use?

Half-Amp
"threshold" detection
the fastest, but mistakes noise spikes for events
works with Model#Channel Count >= 1
available as an option for the "Idl" button
for stats, make sure the "compute stats" option is selected
SKM
segmental K means finds the most likely (Viterbi) event sequence using a model
reasonably fast and accurate
works with Model#Channel Count >= 1
the default algorithm for the "Idl" button
Amp
Baum-Welch (max gamma) to find the most likely event sequence using a model
slower than SKM
segment size limited by available memory
mainly used on small selections for accurate amplitude estimates
Idl/Base
combines Viterbi or Baum-Welch idealization with Kalman baseline tracking
as event detection in the presence of baseline drift is nearly impossible, Idl/Base is
  • slower than the rest
  • sensitive to parameters such as baseline drift std.deviation
  • sensitive to correlated noise
Stat
not an idealizer -- it computes statistics for data that is already idealized
Stat is called internally by the other idealizers

Idealization stats are displayed in the Results window (View -> Results). If your data is segmented, or you are working on a selection list, stats and amplitude histograms are generated for each segment.

The Idealization Process

First, prepare your data. Correct baseline drift if any, unless you're using Idl/Base. Decide if you need a filter ("Fc" at the top) -- for good data it shouldn't be necessary and may distort transitions.

Next, prepare a model. A simple two-state model is usually enough. Set rate constants that match what you see in the data (exit rate = 1 / mean lifetime). Initialize the amplitudes by selecting a sample of single-channel activity, right-clicking the model and "grab all amps." Alternatively, you can select a sample of each level, right-click a state and "grab." Increase the channel count if there are multiple channels in the patch.

Idealize a short selection of data. Check the idealization in the Data window. Does it match the data? Check the histogram in the Results window. Does the fit curve match the histogram? Do the colored component curves have distinct means?

If something is not right:

  • Try turning re-estimation off or on.
  • Try fixing the amplitudes or noise of one or more conductance classes (in Idealization properties)
  • Try different rates, or a more elaborate model. You may need separate states for long and short closures
  • Try a different idealizer
  • Try changing the filter

When you're satisfied, idealize the whole file.

Tutorials


Prev: Tutorial:Simulation Outline Next: Tutorial:Single Channel Measurements
Personal tools