HarmonyCell: Automating Single-Cell Perturbation Modeling under Semantic and Distribution Shifts

TL;DR

HarmonyCell is an automated framework that addresses semantic and distribution heterogeneity in single-cell perturbation studies, enabling scalable, dataset-agnostic modeling with high validity and competitive performance.

Contribution

It introduces a novel end-to-end agent combining LLM-driven metadata unification and adaptive architecture search to handle heterogeneity in single-cell data analysis.

Findings

Achieves 95% valid execution rate on diverse datasets.

Outperforms general agents in out-of-distribution tests.

Matches or exceeds expert-designed baselines.

Abstract

Single-cell perturbation studies face dual heterogeneity bottlenecks: (i) semantic heterogeneity--identical biological concepts encoded under incompatible metadata schemas across datasets; and (ii) statistical heterogeneity--distribution shifts from biological variation demanding dataset-specific inductive biases. We propose HarmonyCell, an end-to-end agent framework resolving each challenge through a dedicated mechanism: an LLM-driven Semantic Unifier autonomously maps disparate metadata into a canonical interface without manual intervention; and an adaptive Monte Carlo Tree Search engine operates over a hierarchical action space to synthesize architectures with optimal statistical inductive biases for distribution shifts. Evaluated across diverse perturbation tasks under both semantic and distribution shifts, HarmonyCell achieves a 95% valid execution rate on heterogeneous input datasets (versus 0% for general agents) while matching or even exceeding expert-designed baselines in rigorous out-of-distribution evaluations. This dual-track orchestration enables scalable automatic virtual cell modeling without dataset-specific engineering.