Skip to main content

Using Networks to Characterize Biological Behavior

In the first few lectures of this course we’ve discussed the relevance of concepts such as strong triadic closure, auctions and game theory, and how these concepts are applied primarily to social networks. Although analyzing social networks is very important and a good example for some of these core concepts, I’d like to pump the brakes a little bit with social networks and talk about some applications of networks people may not be aware of.

As we know, human cells are very complicated and it has taken a long time to get a basic understanding of them. It takes a lot of time and money to perform various biological experiments including techniques such as DNA sequencing, PCR, cell cultures etc. however this approach is essential for characterizing in vivo behavior of cells and their components. This basic knowledge can be applied to beginning to understand diseases such as cancer, providing a framework for how a normal cell should behave, and analyzing how cancer cells deviate from normality. With this framework, is there a way to potentially predict the behavior of a cancer cell to run experiments to aid development of anti-cancer therapeutics? The answer is: yes, with signal transduction pathway modeling.

In the attached article by Ryan Tasseff, a recent Cornell University PhD graduate, he describes how a network model of prostate cancer helps identify pathways toward Androgen Receptor activation, and how this fuels tumor growth in androgen deficient cancer cells. The network was constructed by initially writing down all of the interactions between all genes, proteins, and machinery known in prostate cancer via literature. These interactions were then modeled with differential equations based on mass-action kinetics with rate constants corresponding to each interaction. A way to relate this to what people might be familiar in class is, the larger the rate constant is, the stronger the interaction between two species (or nodes) in the network is. Using many statistical and computer programming techniques, the model was trained against actual experimental data to have the network respond to stimuli analogously to a living prostate cancer cell. For example, if you add DHT to a prostate cancer cell you will see an increase in PSA; the computer modeled network will characterize this behavior – meaning if you change the amount of DHT in the model after it has run to steady state, it will increase the expression of PSA. Using the network to identify critical pathways where the signal transduction is strong is important because one of these critical pathways may not have been identified before, and could be a potential target for anti-cancer therapeutics.

Yes, this example is a very complex one…however, real-world problems are rarely trivial. Networks that characterize cancer cell behavior illustrate the strength and significance of various pathways using similar, yet more complex concepts used in how strength of friendships in social networks characterize who might be a good friend for a social network to suggest to you.

The link to the paper can be found here, including an option for a PDF of the puplication:

The reading may be a bit dense for those not biologically/mathematically inclined, however, the core concept of using a network to describe tumor behavior is very interesting.

For more reading into networks in other fields, take a gander at this article on work done by MIT on potential city planning uses for networks:


Leave a Reply

Blogging Calendar

September 2011
« Aug   Oct »