Understanding protein interactions and pathway knowledge is essential for comprehending living systems and investigating the mechanisms underlying various biological functions and complex diseases. While numerous databases curate such biological data obtained from literature and other sources, they are not comprehensive and require considerable effort to maintain. One mitigation strategies can be utilizing large language models to automatically extract biological information and explore their potential in life science research. This study presents an initial investigation of the efficacy of utilizing a large language model, Galactica in life science research by assessing its performance on tasks involving protein interactions, pathways, and gene regulatory relation recognition. The paper details the results obtained from the model evaluation, highlights the findings, and discusses the opportunities and challenges.

We present 1) a work in progress method to visually segment key regions of scientific articles using an object detection technique augmented with contextual features, and 2) a novel dataset of region-labeled articles. A continuing challenge in scientific literature mining is the difficulty of consistently extracting high-quality text from formatted PDFs. To address this, we adapt the object-detection technique Faster R-CNN for document layout detection, incorporating contextual information that leverages the inherently localized nature of article contents to improve the region detection performance. Due to the limited availability of high-quality region-labels for scientific articles, we also contribute a novel dataset of region annotations, the first version of which covers 9 region classes and 822 article pages. Initial experimental results demonstrate a 23.9% absolute improvement in mean average precision over the baseline model by incorporating contextual features, and a processing speed 14x faster than a text-based technique. Ongoing work on further improvements is also discussed.