Scientific research trends and interests evolve over time. The ability to identify and forecast these trends is vital for educational institutions, practitioners, investors, and funding organizations. In this study, we predict future trends in scientific publications using heterogeneous public sources, including historical publications from PubMed, research and review articles, and patents. We demonstrate that scientific trends can be predicted five years in advance, with preceding publications and future patents serving as leading indicators for emerging scientific topics. We found that the ratio of reviews to original research articles is an informative feature for identifying increasing or declining topics, with declining topics having an excess of reviews. We find that language models provide improved insights and predictions into topic temporal dynamics. Our findings suggest that similar dynamics apply to molecular, technological, and conceptual topics across biomedical research.

翻译：科学研究趋势与关注点随时间的推移而不断演变。识别并预测这些趋势对于教育机构、从业者、投资者以及资助组织至关重要。在本研究中，我们利用包括PubMed的历史论文、研究综述文章及专利在内的多种公开来源数据，预测未来科学出版物的趋势。我们证明，科研趋势可提前五年进行预测，且已有论文与未来专利可作为新兴科学主题的领先指标。研究发现，综述文章与原创研究论文的比例是识别上升或下降主题的有效特征，其中下降主题的综述比例偏高。同时，语言模型能更深入地揭示主题的时间动态特性，从而提升预测能力。我们的研究结果表明，生物医学研究中的分子、技术及概念主题均呈现类似动态规律。