Back

ChromSkills enables interpretable and reproducible agentic chromatin data analysis

Zhang, Y.; Wang, Y.; Tan, Y.; Zhang, Y.

2026-02-06 bioinformatics
10.64898/2026.02.04.703888 bioRxiv
Show abstract

High-throughput assays generate diverse chromatin datasets that require flexible workflows and context-dependent parameter choices. Although large language models (LLMs) can assist analysis, unconstrained LLM-based execution often exhibits unstable behavior and limited reproducibility. We present ChromSkills, a curated library of domain-specific analytical Skills for agentic chromatin data analysis on the Claude Code platform. ChromSkills encodes expert-informed decision logic and parameter-selection rules as modular, human-readable Skills and couples them with structured tool interfaces to enable reproducible execution and interpretable workflow composition from natural-language task descriptions. Across representative chromatin analysis tasks, ChromSkills consistently selected appropriate tools and parameters across repeated runs and improved execution stability and token efficiency when using Model Context Protocol-based tools. Together, ChromSkills provides a practical framework for trustworthy AI-enabled, agentic chromatin data analysis by guiding LLMs with transparent, reusable domain knowledge.

Matching journals

The top 4 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.