<b>Dataset for </b><b>Interpretable Machine Learning Reveals Structural Insights into Selective C–H Borylation by Metal-Organic Framework-Supported Ni Catalysts</b>

Metal-organic frameworks (MOFs) offer a vast design space as tunable supports for catalytic applications, yet their structural complexity often obscures structure-activity relationships. This study investigates MOF-supported nickel (Ni) catalysts for selective sp³ and sp² C–H borylation. Using interpretable machine learning, we developed 45 concise and comprehensive descriptors that capture the diverse structural features of the vast MOF family, derived from over 470,000 MOF structures. These descriptors allowed us to identify key factors governing the sp³ vs. sp² selectivity. Our findings reveal that sp³ C–H borylation occurs within MOF pores via radical-mediated hydrogen atom transfer (HAT), while sp² C–H borylation is associated with surface or defect sites, favoring a concerted metalation-deprotonation (CMD) mechanism. Guided by these insights, we designed Ni catalysts achieving sp³ C–H selectivity of up to 97.8% and sp² C–H selectivity of up to 88.7%. This work provides a systematic framework for rational design of and transferable insights into MOF-supported catalysts.金属有机框架（Metal-organic frameworks, MOFs）作为可调谐的催化载体具备广阔的设计空间，但其结构复杂性往往会掩盖构效关系（structure-activity relationships）。本研究针对用于选择性sp³与sp²碳氢键硼化反应的MOF负载镍（nickel, Ni）催化剂展开探究。研究团队借助可解释机器学习（interpretable machine learning），基于超过47万个MOF结构，开发了45个简洁且全面的描述符，用以捕捉庞大MOF家族的多样结构特征。通过这些描述符，我们明确了调控sp³与sp²反应选择性的核心因素。研究发现，sp³碳氢键硼化反应发生于MOF孔道内部，通过自由基介导的氢原子转移（radical-mediated hydrogen atom transfer, HAT）机制进行；而sp²碳氢键硼化反应则与MOF表面或缺陷位点相关，更倾向于协同金属化去质子化（concerted metalation-deprotonation, CMD）机制。基于上述研究洞察，我们设计的镍催化剂实现了最高97.8%的sp³碳氢键选择性与最高88.7%的sp²碳氢键选择性。本研究为MOF负载催化剂的合理设计提供了系统性框架，同时为相关领域提供了可迁移的理论见解。