Expert opinions in cancer metastasis: Harvesting knowledge from uncertainty and discrepancies

Summary form only given. Research in computational biology is often contingent on principal notions. Mathematical modeling is relying on valid initial assumptions. Text mining algorithms can only retrieve or extract information found in text. Knowledge representation requires a degree of knowledge consensus. Our understanding of certain areas in biology, however, is still in its infancy having a ripple effect in computational efforts. In this talk we discuss a study on cancer metastasis - a complex biological phenomenon with vast clinical importance. Individual viewpoints from 28 experts in clinical or molecular aspects of cancer metastasis were harvested and summarized computationally. Detailed analysis of the data reveals areas of disagreement and a range of opinions on underlying causes and processes in metastasis. The language that experts used while communicating their views was also examined. The experts use gripping metaphors and much hedging. Extensive automatic analysis reveals high use of language associated with cognitive processes (certainly and insight, in particular) language commonly under-represented in scientific text. The results from this study show that in reality knowledge is not as crisp as the view one might obtain by looking at textbooks and the scientific literature. There is speculation, uncertainty and difference of opinion. These findings have ramifications in (i) building mathematical models of biological processes such as cancer metastasis, and (ii) formally representing metastasis. We propose probabilistic models and ontologies that systematically factor experts´ hunches and speculations. We will also discuss the repercussions of this difference of opinion in scientific paper and grant reviewing.