BIOGEN: Evidence-Grounded Multi-Agent Reasoning Framework for Transcriptomic Interpretation in Antimicrobial Resistance
Summary: arXiv:2510.16082v5 Announce Type: replace-cross
Abstract
Interpreting gene clusters from RNA sequencing (RNA-seq) remains challenging, especially in antimicrobial resistance studies where mechanistic insight is important for hypothesis generation. Existing pathway enrichment methods can summarize co-expressed modules, but they often provide limited cluster-specific explanations and weak connections to supporting literature.
Introduction
In recent years, the rise of antimicrobial resistance (AMR) has posed significant challenges to public health, prompting researchers to seek deeper insights into the genetic underpinnings of this phenomenon. RNA sequencing has emerged as a vital tool for studying gene expression, yet the interpretation of RNA-seq data, particularly in the context of AMR, has proven to be a formidable task.
Introducing BIOGEN
To address these challenges, we present BIOGEN, an evidence-grounded multi-agent framework designed for post hoc interpretation of RNA-seq transcriptional modules. This innovative approach combines:
- Biomedical Retrieval: Efficiently sourcing relevant literature and biological data.
- Structured Reasoning: Utilizing a systematic method for deriving conclusions from the retrieved data.
- Multi-Critic Verification: Ensuring accuracy and reliability through multiple evaluation criteria.
BIOGEN aims to generate traceable cluster-level explanations with explicit evidence and confidence labels, enhancing the interpretability of RNA-seq data.
Performance Insights
In a primary study involving a Salmonella enterica dataset, BIOGEN demonstrated exceptional performance, achieving the following metrics:
- BERTScore: 0.689
- Semantic Alignment Score: 0.715
- KEGG Functional Similarity: 0.342
- Hallucination Rate: 0.000 compared to 0.100 for an LLM-only baseline
Furthermore, across four additional bacterial RNA-seq datasets, BIOGEN maintained a zero hallucination rate under the same fixed pipeline, underscoring its robustness in various contexts.
Conclusion
Comparative analyses with representative open-source agentic AI baselines revealed that BIOGEN was the only framework that consistently preserved zero hallucination across all five datasets evaluated. These findings suggest that reliance on retrieval alone is insufficient for achieving reliable biological interpretation. Instead, BIOGEN’s evidence-grounded orchestration represents a significant advance in transparent and source-traceable transcriptomic reasoning, paving the way for future research in antimicrobial resistance.