Matching Cancers With the Best Chemical Treatments

Roland Piquepaille writes "When oncologists meet a new patient affected by a cancer, they have to take decisions about the best possible treatment. Now, U.S. researchers have devised an algorithm which matches tumor profiles to best treatments. They've used a panel of 60 diverse human cancer cell lines from the National Cancer Institute — called NCI-60 — to develop their "coexpression extrapolation (COXEN) system." As said one researcher, "we believe we have found an effective way to personalize cancer therapy." Preliminary results have been encouraging and clinical trials are now planned."

Comments

Re:Insurance

[Daniel Dvorkin]

The application of the algorithm will come well after the "full examination by a professional" stage -- they'll be using it once the cancer has been diagnosed, and they're deciding on which of several specific treatments to use.

Re:Insurance

[piojo]

Lets just hope that doctors who use this algorithm still throughly examine every patient before beginning treatment, because, while probably useful, I doubt its as effective as a full examination by a professional. Actually, if I remember correctly, an algorithm is better than doctors at diagnosing heart attacks... something about doctors being too human, and being unable to ignore statistically unimportant factors such as age (that is, being younger makes you less likely to experience a heart attack in just the same way that being younger makes you less likely to experience the symptoms of a heart attack--a given set of symptoms is equally to indicate a heart attack, regardless of age). My source? Blink, by Malcolm Gladwell. I'm probably misremembering a some of the details, but the point is there.

You may be right about its effectiveness in some cases, but its correctness, once it's perfected, will most likely be statistically better than the judgement of doctors.

Re:Doctors generally won't like this

[macklin01]

Don't forget that the gap needs to be bridged from both sides: while it will indeed take some cultural changes in the medical community to use computational / predictive tools in choosing therapy, it will also require cultural changes in the modeling community to facilitate this. Furthermore, doctors' trust in computational tools must be earned by a well-validated track record of results by the mathematical / engineering community. Interestingly, these cultural changes are underway and can already be observed.

My primary field of research is developing computational tools for modeling cancer progression and angiogenesis, primarily using a PDE point of view where I model nutrient transport within the body and uptake by tumor cells, some simple biomechanics, the degradation and remodeling of the extracellular matrix by the tumor, and the resulting motion of the tumor boundary within the tumor. In fact, this was my dissertation topic just a little over a month ago; the interested reader can see my publications here and some animations of cancer simulations here.

In the several years I've been doing this work, I've seen interesting changes on both sides of the aisle. The mathematical models of cancer have grown in sophistication and realism at an incredible speed. Five or six years ago, models would only examine a single, isolated aspect of cancer growing in homogeneous tissues that were more idealized than even simulated in vitro petri dishes; today, they model many aspects of cancer and the interaction between those aspects. Several years ago, the models were little more than interesting mathematical objects with simplified, spherical solutions that weren't very interesting outside the mathematical community; today, we're simulating complex tumor shapes in fairly realistic tissues, and the results are shedding light on current problems in cancer biology that are otherwise difficult to understand.

Several years ago, it was difficult to even get doctors, oncologists, and others to even look at our research (in our field in general). Today, we're building a track record of results that makes the work easier to trust. Mathematicians and engineers are also realizing the need to acquire the "vocabulary" and biological background necessary to communicate with doctors and biologists, and they're making moves to bridge the gap and collaborate. In the meantime, more cancer biologists are realizing that it takes more than studying isolated cells to understand cancer systems, and they're reaching out to mathematicians to model these complex systems.

The result: very rich and exciting collaborations between doctors and mathematicians to develop helpful predictive tools. My group (at the UT Health Science Center in Houston, with the M.D. Anderson Cancer Center) is doing exciting joint work with oncologists, biologists, mathematicians, and engineers to combine experiments with well-calibrated models of glioblastoma, an aggressive form of brain cancer. Sandy Anderson and Vito Quarnata are doing similar joint mathematical/biological work on breast cancer at Vanderbilt and the University of Dundee, and their work has been featured on slashdot before.

So, it really requires growth toward collaboration from both sides, but fortunately, the need for this has been recognized by both communities and is occurring as we speak. It's a very exciting time in cancer systems biology and computational / predictive oncology! -- Paul