BioInformatics Agent (BIA): Unleashing the Power of Large Language Models to Reshape Bioinformatics Workflow

Bioinformatics plays a crucial role in understanding biological phenomena, yet the exponential growth of biological data and rapid technological advancements have heightened the barriers to in-depth exploration of this domain. Thereby, we propose Bio-Informatics Agent (BIA), an intelligent agent leveraging Large Language Models (LLMs) technology, to facilitate autonomous bioinformatic analysis through natural language. The primary functionalities of BIA encompass extraction and processing of raw data and metadata, querying both locally deployed and public databases for information. It further undertakes the formulation of workflow designs, generates executable code, and delivers comprehensive reports. Focused on the single-cell RNA sequencing (scRNA-seq) data, this paper demonstrates BIAs remarkable proficiency in information processing and analysis, as well as executing sophisticated tasks and interactions. Additionally, we analyzed failed executions from the agent and demonstrate prospective enhancement strategies including selfrefinement and domain adaptation. The future outlook includes expanding BIAs practical implementations across multi-omics data, to alleviating the workload burden for the bioinformatics community and empowering more profound investigations into the mysteries of life sciences. BIA is available at: https://github.com/biagent-dev/biagent.