Computer Simulation of Cancer Growth

Roland Piquepaille writes "For a long time now, researchers and scientists have used computer simulations in the physical sciences: physics, chemistry, and engineering. But what about biology? An international team of U.S. and Scottish mathematicians and biologists has built a math model to predict tumor behavior. The researchers say their approach is similar to the one used by weather forecasters. So far, this approach is entirely theoretical. But the scientists see their effort as the beginning of a new era in cancer research — 'a sea change in how biology is being done,' as the lead researcher described it. Read more for additional references and illustrations about this use of computer simulation to predict a cancer evolution."

Comments

Couldnt these....

[Creepy Crawler]

Programs and techniques be used wherever chaotic systems take place? I guess it's in the domain of the weather, disease rates and population growth.

It would be rather interesting to watch social networks in the similar style (Im not thinking of myspace gunk...).

Connections with Stem Cells?

[mcrbids]

Increasingly, researches seem to be finding a clear connection between stem cells, aging, and cancer. It looks like cancer depends on errant stem cells for its rejuvination - and years of cancer study supports this theory.

So, by all appearances, if we could destroy just the right cells, a small percentage (0.10%) of the tumor, the tumor goes away!

So, while the mathematical model of growth might represent some predictive value, it certainly will not effectively model new developments, such as the above, when they are found.

Re:Connections with Stem Cells?

[macklin01]

So, while the mathematical model of growth might represent some predictive value, it certainly will not effectively model new developments, such as the above, when they are found.

There's still plenty of value to be found in higher-scale models. (e.g., how the tumor as a whole interacts with the microenvironment, how proliferation-induced pressure turns off the vasculature and prevents drug delivery, how oxygen and glucose delivery throughout the tumor and the microenvironment affects the patterning of hypoxic and necrotic cells, which, in turn, affects angiogenesis and matrix degradation) Cancer is a multiscale problem, with interaction between all the scales. Focusing on one scale alone (molecular or tissue-scale) likely will not solve the entire problem.

In fact, developing a good tissue-scale model is a natural step toward creating a multiscale model, where molecular- and cell-scale dynamics affect the growth parameters that govern tissue-scale behavior. (Similarly to how the behavior of individual molecules leads to things that can be averaged at larger scales, like heat, viscosity, etc.) First, you fix the parameters and neglect the small-scale dynamics to figure out the large-scale behavior. Then, you model the small-scale dynamics and learn how to couple them to the previously-fixed parameters.

On a tangent, I also worked with "Sandy" Anderson in a different U.S.-Scottish collaboration. He's a really great guy, as well as Mark Chaplain and Steven McDougall. And Sandy is a pretty incredible cook! :-) -- Paul

Cancer, for the troll

[tempest69]

There are beauties to a computer model. it can perturbed. Lets say that you model a carcinoma, that is undergoing a drug treatment that impairs its ability to build mitochondria.. The Cancer will mutate to switch to a glycolosis process for most of its energy. But to go glycolytic, it will need to grow more capillaries, so you could add in chemicals that prevent angiogenesis. Or you could starve the patient of glucose, providing a super low carb diet, to cause the cancer to die from lack of nutrients.

Basically an oncologist should be able to get a rough guess of how a series of treatments will work, and if a set of treatments is just going to make a more resilient cancer, then they can consider more viable options. Cancer is tricky, and some treatments arent effective at the same stages.

Storm

Cancer as a chaotic system

[MillionthMonkey]

Who said cancer was a chaotic system? Do very small changes in input parameters cause exponentially large deviations in output values? I doubt it. Cancer is probably difficult to simulate due to its complexity, not its chaoticness.

Cancer is complex because it is chaotic. There is a (poorly understood) mechanism cells use to ensure that each chromosome has the right copy number (usu. 2) and that no chromosome is missing a chunk. In a tumor, or even a precancerous growth, something has happened to break this mechanism. When cells in a tumor or precancerous growth divide, they don't copy their chromosomes correctly- they make way more mistakes than normal cells during mitosis. A cell division event might give one daughter cell a single copy of chromosome 5, while the other daughter cell gets three copies. Or, one daughter cell gets a chunk of a chromosome that is missing in the other daughter cell. As a result there is enormous genetic variability within a single tumor or precancerous growth- each one is a little version of evolution in action as individual cells within the tumor population compete for fitness. (Most genes still work if you only have one copy, or 3, but it's much better to have 2.)

Eventually one cell has zero copies of at least a part of a chromosome, and that's when the fun really starts. One of the arms of chromosome 3, for example, appears to confer certain "superpowers" on any cell that loses it, since there appear to be certain tumor suppressor genes on that chromosome. As chromosome parts are gradually lost in the tumor population, the various superpowers of cancer become evident: growth in absence of any growth signals, loss of contact inhibition (you keep dividing even when you run out of room), the ability to ignore suicide signals from attacking white blood cells, the ability to promote blood vessel growth into the tumor, the ability to metastasize, etc. If a cell loses the right chunk of the right chromosome it can quickly take over the entire tumor, and you have a population of cells that are all missing that chromosome chunk and are ready to start losing more random chunks. So as you see, "very small changes in input parameters cause exponentially large deviations in output values".

I could be wrong but I think what they are modeling here is the genetic variation within the tumor, as evident in the chromosomal copy number within each cell.

Re:Cancer as a chaotic system

[Anonymous Coward]

I guess I have to be careful because I was talking about the mathematical definition of chaos which is much more important to computer systems. Having a setpoint of initial conditions between where you have a cancer and don't doesn't necessarily make something chaotic. For a large distribution of cells this can be modeled statistically.

An example of what people generally mean when they say something is chaotic is that say for example you have an ideal population of antelope. If the temperature profiles for two years were very similar you would expect a similar growing season and a similar population level in an ideal situation. If one year had a slightly better input parameter for temperature (for example, higher temperature) that yielded a better growing season you might expect a slightly better population level. And if you increased the parameter again by a very slight amount you would expect an even better growing season. What chaos does is that you can't use simple logic. For each very small change it might be beneficial or detrimental. And even if you half-split it will still be beneficial or detrimental. This doesn't mean that it can't be simulated, but it requires extreme computing power. In our first non-chaotic example if you increase the temperature by 0.1 C you might have a 2% population increase. In our chaotic system if you increase the temperature by 0.00000001 C you might have a 2% population increase about a very small range of temperature (hence the exponential nature)--a further 0.00000001 C increase might mean a 10% decrease.

In reality, chaotic systems are fairly rare because they require non-linear systems that behave in particular ways. Most non-linear systems can be modeled in fairly simple ways and don't break down (or are rarely operated in the chaotic regions). I doubt that cancer is a chaotic system because I'm not convinced that small changes in input parameters will cause dramatic changes in tumors. It appears to me that an enormous part of the progression of a cancer is determined around where it started and how close it is to blood supply and other tissue that might cut it off. The purpose of modeling cancer isn't to see where it would start, but instead to say how it will progress after it is started. If a cancer is next to a blood vessel, I don't see it behaving chaotically. It will explode in growth and be affected in a non-chaotic way by variations in parameters. However, only in the case of a completely isolated cancer with low blood flow can I see the possibility of chaos. But I'm still not convinced that it isn't a threshold in the parameters instead of chaotic behavior that allows isolated cancers to die or progress in an almost random fashion.

Re:Or Not

[pimpimpim]

take the uterus out and you're most likely cured

Your comment is really insightful, but it also reminds me how some doctors treat their patients as an engineer would treat a car. It must be really an unbelievable sad thing to happen. Then again, doctors can't cry over every patient, it would probably kill their spirit.

Ontopic: I quickly read the article, it seems that they especially focus on what happens if cells at certain positions in the tumor are being attacked by treatment. Depending on the type, the more actively replicating cells may be localized at the outside or something (didn't really get that). As they can go over many different schemes in a short time, their research might help optimizing treatment (if lower doses of drugs can be used that will always be better). So it might look straightforward, but this is actually a nice bit of research, done with simple means, that makes it rather elegant I think.