Foundation models (FMs) have exhibited remarkable performance across a wide range of downstream tasks in various domains. Nevertheless, general-purpose FMs often face challenges when confronted with domain-specific problems, due to their limited access to the proprietary training data in a particular domain. In life science, a large portion of biomedical knowledge is encapsulated within heterogeneous biological data, such as molecular structures, wet-lab experiment results, and knowledge bases, which exhibit significant modality gaps with human natural language. In this paper, we introduce BioMedGPT, an open multimodal generative pre-trained transformer (GPT) for biomedicine, to bridge the gap between biomedical data and natural language. BioMedGPT allows users to easily "communicate" with various biological modalities through free text, which is the first of its kind. BioMedGPT aligns different biological modalities with the text modality via a large generative language model, namely, BioMedGPT-LM. We publish BioMedGPT-10B, which unifies the feature spaces of molecules, proteins, and natural language via encoding and alignment. Through fine-tuning, BioMedGPT-10B outperforms or is on par with human and significantly larger general-purpose foundation models on the biomedical QA task. It also demonstrates promising performance in the molecule QA and protein QA tasks, which could greatly accelerate the discovery of new drugs and therapeutic targets. In addition, BioMedGPT-LM-7B is the first large generative language model based on Llama2 in the biomedical domain, therefore is commercial friendly. Both BioMedGPT-10B and BioMedGPT-LM-7B are open-sourced to the research community. We also publish the datasets that are meticulously curated for the alignment of multi-modalities, i.e., PubChemQA and UniProtQA. All the models, codes, and datasets are publicly available.

翻译：基础模型（FMs）已在多个领域的下游任务中展现出卓越性能。然而，通用基础模型在应对特定领域问题时往往面临挑战，因其对特定领域专有训练数据的访问受限。在生命科学领域，大量生物医学知识蕴含于异构生物数据中，例如分子结构、湿实验成果及知识库，这些数据与人类自然语言之间存在显著模态差异。本文提出BioMedGPT——一种面向生物医学的开放多模态生成式预训练Transformer，旨在弥合生物医学数据与自然语言之间的鸿沟。BioMedGPT允许用户通过自由文本轻松"沟通"多种生物模态，这在该领域尚属首次。该模型通过大型生成语言模型BioMedGPT-LM将不同生物模态与文本模态对齐。我们发布了BioMedGPT-10B，通过编码和对齐统一了分子、蛋白质和自然语言的特征空间。通过微调，BioMedGPT-10B在生物医学问答任务上优于或持平于人类及显著更大的通用基础模型。该模型在分子问答和蛋白质问答任务中也展现出优异性能，有望极大加速新药及治疗靶点的发现。此外，BioMedGPT-LM-7B是生物医学领域首个基于Llama2的大型生成语言模型，因此具备商业友好性。BioMedGPT-10B与BioMedGPT-LM-7B均向科研社区开源。我们还发布了为多模态对齐精心整理的数据集，即PubChemQA和UniProtQA。所有模型、代码及数据集均已公开。