Hidden Markov Models for Molecular Motors: ModelChange(M1,M0) - File Exchange

Code covered by the BSD License

Highlights from
Hidden Markov Models for Molecular Motors

DisplayModel(M, L, iterations... Display the HMM described by M, the log likelihood value,
ForBackF(M,Y) Forward-backward algorithm for step-HMM
ForwardBackward_3(M,Y,I0,test... function [LL newM gamma] = ForwardBackward_3(M,Y,I0,test,BigB)
M=MakeMonotonicModel(nu, yQua... function M=MakeMonotonicModel(nu, yQuantum, noiseSigma, transProb, stepSizes, stepSigma)
ModelChange(M1,M0) Check the changes in the transition probabilities. Looks at the
PlotAndLabelPeaks(x,y,minsize... function PlotAndLabelPeaks(x,y,minsize,xscale)
RestorationPlot(Y, EstY, wrap... function RestorationPlot(Y, EstY, wrap, yquantum, EYbase,markerstr) Fancy
WtConvol(a,b,m)
[EstY EstI LP Wrap]=ViterbiRe... % Viterbi algorithm for step-HMM
[X,Y]=StepSimulator(M,N) function [X,Y]=StepSimulator(M,N)
[X,Y]=StepSimulatorC(M,N) Given the number of time points N and the model M, simulate a staircase
[sigma nu]=EstSigmaAndNu(Y,I0) Find reasonable values for these variables based on the step data vector
b=Makeb(nu,sigma,DutyCycle,Er...
ms=QuickShift(m,shift) Faster replacement for circshift for time-critical applications.
ms=QuickShiftMex(m,shift) Faster replacement for circshift for time-critical applications.
n=NextNiceNumber(x,f) function n=NextNiceNumber(x)
xs=Step125(x)
MotorHMM1.m
MotorHMM2.m
View all files

from Hidden Markov Models for Molecular Motors by Fred Sigworth
A set of functions for analysing noisy recordings of the random stepping of molecular motors

ModelChange(M1,M0)

function RelDif = ModelChange(M1,M0)
% Check the changes in the transition probabilities.  Looks at the
% change in N.C(0,i,j) and the change in M.C(~0,i,j)
% relative to the maximum nonzero element.  The maximum of all these is
% returned.
epsi=.001;
[nu ns ns1]=size(M1.C);
A0=squeeze(M0.C(1,:,:));
A1=squeeze(M1.C(1,:,:));
Ad=max(max((abs(2*(A1-A0)/(A1+A0+epsi)))));
% Now we zero out the no-step entry and compute the rest.
M0.C(1,:,:)=0;
M1.C(1,:,:)=0;
N0=squeeze(sum(M0.C));
N1=squeeze(sum(M1.C));
Nd=max(max( squeeze(max(2*abs(M1.C-M0.C)))./(N1+N0+epsi) ));
% Ad
% Nd
RelDif=max(Ad,Nd);

Highlights from Hidden Markov Models for Molecular Motors

Highlights from
Hidden Markov Models for Molecular Motors