Literature Data Mining for Biology

In this introduction, we summarize the papers included in the session on Literature Data Mining for Biology. We then discuss the need for a challenge evaluation for this eld, and the steps to create such an evaluation. These include creating a shared infrastructure, providing annotated data, and deening and implementing common evaluation metrics. This would enable researchers to compare diiering methods, in order to accelerate progress in this eld. In this context, we describe two speciic applications: extraction of biological pathways from the literature and automated database curation. For each of these, we outline the task deenition, the creation of an annotated corpus, and evaluation metrics. Even though the number and the size of sequence databases are growing rapidly, most new information relevant to biology research is still recorded as free text in journal articles and in comment elds of databases like the Gen-Bank feature table annotations. As biomedical research enters the post-genome era, new kinds of databases that contain information beyond simple sequences are needed, for example, information on cellular localization, protein-protein interactions, gene regulation and the context of these interactions. The forerunners of such databases include KEGG 1 , DIP 2 , BIND 3 , among others. Such databases are still small in size and are largely hand curated. A factor that can accelerate their growth is the development of reliable literature data mining technologies. This year is the third time the Paciic Symposium on Biocomputing has devoted an entire session to natural language processing and information extraction for biology. Compared to the last two years, the eld has made tremendous