AblateCell: A Reproduce-then-Ablate Agent for Virtual Cell Repositories

2026-04-21

AI Summary

This study addresses the challenge of systematically attributing performance gains in existing AI-based virtual cell models, which is hindered by the lack of standardization in biological codebases and their tight coupling with domain-specific data. To overcome this, the work proposes the first scalable and verifiable ablation analysis framework tailored for biological codebases. The framework enables end-to-end reproduction of baseline results through automated environment configuration and dependency resolution, constructs isolated code change graphs, and employs a multi-objective reward-driven adaptive experimental scheduler to facilitate closed-loop ablation studies. Evaluated on three single-cell perturbation prediction codebases, the approach achieves an 88.9% end-to-end workflow success rate—29.9% higher than that of human experts—and a 93.3% accuracy in identifying critical components, outperforming heuristic methods by 53.3%.

Technology Category

• Humans and AI: Other Foundations of Human Computation & AI
• Machine Learning: Calibration & Uncertainty Quantification
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Systematic ablations are essential to attribute performance gains in AI Virtual Cells, yet they are rarely performed because biological repositories are under-standardized and tightly coupled to domain-specific data and formats. While recent coding agents can translate ideas into implementations, they typically stop at producing code and lack a verifier that can reproduce strong baselines and rigorously test which components truly matter. We introduce AblateCell, a reproduce-then-ablate agent for virtual cell repositories that closes this verification gap. AblateCell first reproduces reported baselines end-to-end by auto-configuring environments, resolving dependency and data issues, and rerunning official evaluations while emitting verifiable artifacts. It then conducts closed-loop ablation by generating a graph of isolated repository mutations and adaptively selecting experiments under a reward that trades off performance impact and execution cost. Evaluated on three single-cell perturbation prediction repositories (CPA, GEARS, BioLORD), AblateCell achieves 88.9% (+29.9% to human expert) end-to-end workflow success and 93.3% (+53.3% to heuristic) accuracy in recovering ground-truth critical components. These results enable scalable, repository-grounded verification and attribution directly on biological codebases.

Problem

• systematic ablation
• virtual cells
• reproducibility
• biological repositories
• performance attribution

Innovation

• AblateCell
• reproduce-then-ablate
• virtual cell repositories
• automated ablation
• codebase verification