| I focus on several features of knowledge acquisition in biomedical communities. I use two data types for the analyses: the molecular interaction data extracted from biomedical articles using natural language processing and manual annotations of articles listed in PubMed. My study indicates that high-impact research in biomedical sciences appears when highly popular molecular interactions are well-mixed with new interactions, and when popular biological concepts are combined with unpopular chemicals and drugs. The analysis suggests that knowledge of molecular interactions tends to advance by incremental attachment of newly discovered actions to islands of highly interconnected "old" facts. I discuss how information produced in one field potentially useful in other fields stays in local "knowledge pockets" and give an estimate of the number of interactions locked in recent publications. I present a quantitative analysis of the differences between publication types in biomedical literature and study the popular belief that knowledge production by scientific communities is slowing down. Finally, I describe a tool for visualizing the changing popularity and impact of more than 150,000 biological and chemical topics. |
