Auditing Retrieval-Augmented LLM Hypotheses for Longitudinal Cell Painting Morphology

2026-07-17

AI Summary

This study addresses the challenge of translating subtle, chronic perturbations—such as low-dose-rate radiation—into interpretable biological mechanisms, a task hindered by the lack of auditability in hypotheses generated by large language models (LLMs). To overcome this, the authors propose an evaluation-prioritized, retrieval-augmented framework that integrates morphological differences between treated and control groups, neighboring perturbation samples, pathway context, and literature evidence. By anchoring LLM-generated hypotheses to stable evidence identifiers, the approach yields structured, traceable predictions followed by hierarchical summarization. The work introduces two novel quantitative audit tests: V1 verifies that cited evidence exists within the input, while V2 assesses consistency between predicted biological processes and salient morphological features. Experiments demonstrate zero invalid citations in V1 and show that V2’s morphological compatibility increases with perturbation intensity, correlating positively with independent drift metrics, thereby uncovering adaptive phenotypes involving metabolic reprogramming and proteostatic stress under low-dose-rate conditions.

Abstract

High-content morphological profiling (Cell Painting) yields sensitive, high-dimensional signatures of cellular state, but translating longitudinal morphology trajectories into interpretable biology remains difficult, especially for weak, chronic perturbations such as low-dose-rate ionizing radiation. Large language models (LLMs) can synthesize heterogeneous evidence into biological narratives, yet their scientific use requires quantitative auditing. We present an evaluation-first, retrieval-augmented interpretation framework for longitudinal Cell Painting morphology, applied to a 9-week RPE-1 time course across five dose rates (0.003--6.0 mGy/hr). Week-matched treated-control morphology deltas are combined with retrieved perturbation neighbors, pathway context, and literature evidence through stable evidence identifiers, enabling an LLM to generate structured, evidence-linked hypotheses that are hierarchically summarized while preserving provenance. We introduce two quantitative auditing tests: V1 citation validity, which verifies that cited evidence identifiers exist in the prompt, and V2 proxy-based morphology compatibility, which evaluates consistency between predicted biological processes and the most altered morphology features. In our experiments, V1 detected no invalid evidence references, while V2 showed meaningful morphology compatibility that increased with perturbation strength and was positively associated with an independent morphology drift summary. The framework produces auditable, falsifiable biological hypotheses, including an adaptive phenotype involving metabolic reprogramming and proteostatic stress at lower dose rates (0.003--0.3 mGy/hr). Current limitations include proxy-based evaluation and the lack of ground-truth mechanism labels.

Problem

• longitudinal morphology
• Cell Painting
• retrieval-augmented LLM
• hypothesis auditing
• low-dose radiation

Innovation

• retrieval-augmented generation
• longitudinal morphology
• LLM auditing
• Cell Painting
• evidence-linked hypotheses