<p>Interdisciplinary research is a key driver of scientific innovation, yet how knowledge from different disciplines integrates to address research problems remains understudied. Recent advances in large language models (LLMs) provide new tools for analyzing complex texts and extracting domain knowledge. Using bioinformatics as a case study, this paper explores the potential of LLM-based title/abstract analysis to identify contributing disciplines and their functional roles within interdisciplinary research from titles and abstracts, and compares the results with those derived from the topics of references and enriched cited references (ECR) data. The findings suggest that the performance of LLMs in extracting the contributing disciplines and their roles within interdisciplinary research can be enhanced through stepwise prompt optimization. Compared to the ECR-enhanced reference analysis, the LLM-based method focuses more directly on disciplines closely related to the core content of the study and suggests disciplinary contributions that the ECR-enhanced reference analysis may overlook. In the co-occurrence networks of disciplines, the ECR-enhanced reference analysis reports broader and more diverse integration pathways, while the LLM-based network exhibits a more centralized structure, highlighting the strong connections between core fields such as Biochemistry, Genetics &amp; Molecular Biology and Computer Science. Additionally, the LLM-based method suggests more fine-grained contributions of different disciplines across key components of the research process, while the ECR-enhanced reference analysis tends to capture a broader range of disciplinary contexts involved in different sections of a study. Overall, the findings demonstrate the complementary strengths of LLMs and reference analysis in understanding the process of interdisciplinary knowledge integration. Future work could combine these two approaches to develop a more comprehensive methodological framework.</p>