Probabilistic Gene Nets

Probabilistic Gene Nets
Week 6
General Structure
(later we will outline in detail)
New – to be created
I suggest building ruletree first and testing it with
rule and prob to see if you get the ruletree, pnet
and rnet specified in the notes.
[pnet, rnet]=ruletree(rule,prob)
This is still “rule 1” (one rule), it can
just take two forms, with different
associated probabilities.
Rule 1
Biograph makes Rule Tree from ptree
Building ptree
• Each row and each column in ptree are nodes in
• Initialize ptree. To do this you need to find out, from
your cell rule, how many nodes there will be.
– Build a vector called R, where R(i)=the number of
different forms that rule{i} can take.
• Input probabilities from cell prob into ptree. Each
entry specifies the probability that node (row) will
connect to node (column).
ptree, ctd.
-What does the
cell prob look
-add your ids to
ptree and hand
to biograph!
pnet and rnet
• pnet contains the probabilities of arriving at each
of the final nodes in the ruletree. Find this for our
• You can implement this step-wise in MATLAB:
accumulate a prob cell that contains the
probabilities of arriving at the nodes in each
• For example stepprob={[each node in layer 1]
[each node in layer 2]…
[each node in final layer]}
In this case, what is pnet?
pnet and rnet continued!
rnet is “a vector of N rule indicators”
What does this mean?
For each final node, what is it telling us?
Let’s build by hand for our small example.
How to encode: use powers of 10?
Do same way as pnet: build cell then take final
entry which corresponds with terminal nodes.
The Driver: pbndriver
• Where STM is made, among other things
• What other things?
• Specify wire, rule, prob
• Draw gene net itself just based on wire
• Call ruletree on rule and prob to return pnet and
• Build STM
– Initialize
– Build using pnet and genestm
• Display the STM using imagesc
• Display the STM using biograph
Building STM
• Run last week’s genestm on wire and the
different rule combinations.
• genestm will output an STM. What is the
probability that this STM will occur? Where do
we find it?
• The final STM is simply the sum of all possible
STMs generated by genestm, weighted by the
probabilities that they will occur, found in

similar documents