Current sequencing efforts churn out massive amounts of (missense) mutations for a given reference genome – so cancer-cell lines are an obvious and worthy target for these shiny new big guns. Adressing the question of what these mutations are actually doing (wrong) and how they affect structure and function of proteins is not quite straightforward. The problem is that each tumor has a slightly different set of mutations, even parts of the same tumor might not be genetically identical. Are there detectable trends that would tell us more about the molecular mechanisms involved?
One could speculate about the dominant mode of action in tumor initiaition / progression at the molecular scale: For example, mutating a small, hydrophobic amino-acid in the core into a big, charged one might screw up the entire structure, rendering the protein inactive (loss-of-function). Alternatively, one could imagine how mutations at the surface alter binding affinity / specificity and cause havoc in the downstream signalling and regulatory pathways (gain-of-function).
The figure above illustrates the structural location of mutations found in cancer patients (red spheres) and important sites for the function of this protein. Proximity between mutations and functional sites is often not visible in the sequence, but becomes apparent in the context of the three dimensional structure. The effects of mutations on functional sites provide insights into possible mechanisms of cancer initiation and help to select subsequent experiments.
In this paper, a structural-bioinformatics approach comes to the rescue. As you might have suspected, actually BOTH hypotheses are true. This becomes evident when distinguishing oncogenes and tumor-supressors: taken alltogether, the structural patterns of known missense mutations do not deviate very much from what you’d expect from a random distribution. After stratification into oncogenes and tumor-supressors, a very robust pattern emerges: in oncogenes, gain-of-function mutations dominate, in tumor-supressors loss-of-function mutations are more frequent. This makes a lot of sense – but don’t get fooled by the conceit of hindsight, pretending it was all clear from the outset: “The fact that a hypothesis makes sense does not eliminate the need to test it as rigorously as possible” (found in this article by Robert E. Kingston) is also true here.
The resulting classifier is not very picky about the quality of the underlying structural model, allowing for a robust pre-classification of cancer-associated genes into tumor-supressors and oncogenes based on available sequencing data and the resulting mutation patterns. This contributes towards elucidating the as-yet unclear function of additional cancer-associated genes without the need to perform extensive experiments.
The structural impact of cancer-associated missense mutations in oncogenes and tumor suppressors
Henning Stehr, Seon-Hi J Jang, José M Duarte, Christoph Wierling, Hans Lehrach, Michael Lappe and Bodo MH Lange
Molecular Cancer 2011, 10:54 doi:10.1186/1476-4598-10-54
The electronic version of this article is the complete one and can be found online at: http://www.molecular-cancer.com/content/10/1/54. Thanks to Henning Stehr for figure and text contributions.