A systems approach to cancer
First off I hate cancer. I hate the disease almost as much as the vulturous profiteers who are assosciated with this ubiquitous disease.
That being said I am at the NYAS meeting, “A systems approach to the study of cancer.”
Andrea Califano: Interactome Analysis Reveals Master Regulators of Human Malignancies. ARACNe reverse engineering regulatory networks. Hmm if you apply information theory to regulatory networks you can predict edges to look for. ChIP on chip data has been show to have a false negative rate an order of magnatude below the predicted and later proven interactions.
The. Ext genome scale project should be devoted to assembly and validation of whole regulatory networks.
You definitely have to be a beleiver in Sys Bio to buy into this.
Galit Lahav: Dynamics of the p53 signaling pathway
P53 and Mdm2 regulate each other via transcription and through direct interaction so it works in a slow grade way and through damped oscillations. Sweet they made fusions that report relative levels oscillating and do a 30 hour time lapse movie. There we see they aren’t damped oscillations it’s just that more cells oscillate more under stress. You can show whole mouse p53 oscillations. Perfect for in vivo imaging. So the pulsitile dynamics allow for a wider range of regulation depending on the frequency of oscillation vs the stability of the regulated protein.
In order to look at a sychronized population they image individual cells and retrospectively synch their cycles with image processing. Another common in vivo method repeated.
Chris Sander: Network Pharmacology of Cancer
Anyone who starts off telling you how they failed to solve the protein folding problem is OK by me. Structure is function, but so is function. This is another repeating motif. He takes pairs of drugs to cause network perturbations. Hey this is the first time I recognized a differential equation! Go Hopfield network model!
Oh man I just realized why network modeling actually works. Even if you are missing nodes the model is still somewhat accurate. In a structural model you have to have every atom as a vector even if it’s a chaperone or membrane or whatever random thing the folding protein bumps into. The former bravado of a structural modeler is actually realistic in a network milieu.
Arnold Levine: single nucleotide polymorphisms in the p53 pathway
There are roughly 5 snps per gene in people. Not counting the untranslated regions. 18million total. SNPs and populations getting cancer. It reminds me that we would help more people if we spent our research budgets on public health. The other nice thing about SNPs is you can look at phylogenies over the last 30,000 years. I really appreciate that he presents all sorts of hypotheses about why these SNPs are more or less prevalent.
It turns out p53 upregulates LIF needed for implantation of embryos. So low levels of p53 can prevent pregnancy in women. It also can sense aneuploidy and reject implantation. So p53 can control germline cells. P53 is more of a germline preservation mechanism. The stem cell functions only came later in mammals. There is a p53 in very primitive animals.
This symposium shows a real turning point for me and cancer. We are finally realizing the deeper implications of cancer and stem cells and of course systems modeling as an informant for benchwork. Plus it gave me some solid ideas on dynamic imaging.