T-61.5110 Intensive course on modeling biological networks
Back to the course web page
Project work
-
Topic: choose yourself; can be about the topics of lectures but it
is even better to use your own data. We will give some suggestions
in the exercise sessions, but they are not complete ideas. The topic
should naturally be about modeling biological networks.
-
The work needs to includes some technical development as well as
at least some biology. For example demonstrating real understanding
of some existing method by implementing and studying the properties
of it is sufficient. Just running some simulations with existing
ready-made software is not enough, though experiments designed
to non-trivially compare algorithms would be acceptable. You are
also free to use existing software as part of your work, as long
as it still includes some implementation by yourself.
-
The exercises will provide supporting material; you can get data
from the sources given in the first exercise (or re-use the data
used in some other exercises), and you can continue
the methods from what we did in the other exercises.
-
Groups (2-4 persons) or solo; everybody turns in a separate report.
It is advisable to form groups with different kind of backgrounds
(helps tremendously in both implementation and interpretation),
but remember that everyone should do some of the programming/simulation
work.
-
For groups: name your group members in the report, and point out
your individual contribution.
-
Turn in a 0.5 page plan by the end of this week. These should be sent
to t615110@cis.hut.fi
with subject "Project plan". The subject is important for automatic
filtering, so please make sure it is correct.
-
Deadline for a report: end of August. Computer accounts are up until
the end of the month as well. Again send the report, preferably in
pdf, to t615110@cis.hut.fi,
this time with subject "Project report".
-
Report: free form, about 4 pages written clearly and
concisely. Include all necessary details, in separate appendices if
necessary. You can write the report using either English or Finnish.
-
The report/work will be graded, and the course grade is based solely
on that.
Some possible topics
Below is a list of some of the methods described on the course.
Any of these can be used as a basis for the project work. More
topic ideas will be added when the course proceeds. On Thursday
we will hear at least about Bayes networks, and on Friday more about
genetic regulatory networks. You should find papers about all with
some Google searches, using the names of the methods and the authors
mentioned on the lecture slides. Ask if you don't.
-
Graph-based clusterings (see the clustering exercise); spectral-clustering
-
Mixture models for clustering expression data (might be a bit too simple
and boring...)
-
Bi-clustering of expression, for example plaid models
-
Topic models of expression data (you can for example focus on the
discretization part or the modeling part, or alternatively even
simplify(?) the model by replacing discrete distibutions by Gaussians)
-
Expression programs; regression trees and modules
-
The Gram algorithm
-
Factor graphs; see the exercises of Thursday
-
Supervised link prediction; see lectures of Thursday (Naive Bayes etc)
-
Mixed membership stochastic blockmodels (Thursday)
Printing
See http://www.tkk.fi/cc/services/printing/
for instructions on printing. Printers maari1 and maari2 are in this floor, and
you should be able to select them directly from Acroread.
Reaching the computers from outside
The computer accounts are valid until the end of August. You can take ssh connection to "vipu.hut.fi". That machine, however, does not have Matlab and should not be used for computing. You should thus proceed and take ssh connection from that to any of the machines in classrooms. Here is a list of all the available computers; you should be able to use most of those. The list under "Jalokivet, (Maari-C)" has the computers we were using during the course.
Page maintained by t615110 (at) cis.hut.fi,
last updated Friday, 17-Aug-2007 15:57:16 EEST