Neil Lawrence ML@SITraN

Current release is 0.11.
Release 0.11 contains a bug fix for file chipVarEstepCMu.m
Release 0.1 is the first release associated with the paper, containing scripts for recreating the results given.
The toolboxes required to run the code are listed below.
Toolbox  Version 
NDLUTIL 
0.15 
Finally you will also need the NETLAB toolbox in your path.
demTuVar
and demSpellmanVar
which run the code on the metabolic and cell cycle data sets. Also of
interest may be the script demFakeVar
which
runs the code on artificial data. demTuVar
and demSpellmanVar
invoke the functions chipVarEMmu
and chipVarEM
respectively
which compute the posterior estimates of the latent variables by
maximising the
variational lower bound on the likelihood. The functions used are
different as there
is no need to estimate the baseline expression level for cDNA array
data such as the
cell cycle data set. The scripts output a variable model, containing
the parameter
values which maximise the likelihood, and expectationsB (posterior
expectations of
the regulatory intensities) and expectationsC, posterior expectations
of the protein
concentration profiles. The function chipVarEMmu
also
outputs a variable
expectationsMu containing posterior statistics for the distribution of
the baseline
expression levels.
The variable expectationsC has four fields: .entropy, a number
contatining the posterior
entropy of the approximating distribution, .c, a matrix with as many
rows as transcription factors and as many columns as time points, .ccT,
a 3D array containing
the correlations among transcription factors at each time point, and
.cAltc, a 3D array
containing correlations among transcription factors at consecutive time
points. By specialising the transcription factor index one can obtain
plots of the transcription
factor protein concentrations with errorbars. An example is show below
for the transcription factor ACE2 during the cell cycle.
expectationsB.errorbars
contains the associated error bars. By specifying to a transcription
factor, we can
obtain a distribution of the significance levels of its regulatory
intensities. An example of this is shown below for the transcription
factor ACE2 during the cell cycle.