State Function-Based Correction: A Simple and Efficient Free-Energy Correction Algorithm for Large-Scale Relative Binding Free-Energy Calculations

Free-energy perturbation-based relative binding free-energy (FEP-RBFE) calculations have become an important tool in drug discovery, but inherent computational errors require corrections based on fundamental physical principles to improve prediction accuracy. Traditional correction methods enforce physical consistency by identifying cycles in perturbation graphs, but their computational cost grows exponentially with network size due to the combinatorial explosion of cycles. This severely limits their applicability to modern drug discovery, where large-scale FEP-RBFE screens involving hundreds to thousands of ligands are increasingly common. We present an efficient and straightforward State Function-based Correction (SFC) algorithm, which leverages the state function property of free energy without requiring cycle identification. This eliminates computational bottlenecks, with the computational cost scaling as O(P × N), where P is the number of edges in the perturbation graph and N is the number of molecules. In contrast to graph-based methods such as weighted cycle closure (WCC), SFC maintains consistent computational efficiency across increasing graph sizes, enabling the efficient handling of large perturbation networks with up to 50 000 molecules or even moreuseful for high-throughput FEP-RBFE applications. Furthermore, SFC incorporates uncertainty-aware weighting to further enhance correction performance. These advantages position SFC as an efficient RBFE correction method to better support high-throughput FEP-RBFE calculations aimed at lead optimization.基于自由能微扰的相对结合自由能（FEP-RBFE）计算已成为药物发现领域的关键工具，但其固有的计算误差需基于基本物理原理进行校正，以提升预测精度。传统校正方法通过识别微扰图中的环来确保物理一致性，但受限于环的组合爆炸问题，其计算成本随网络规模呈指数级增长，这极大限制了其在现代药物发现中的应用——当前涉及数百至数千个配体的大规模FEP-RBFE筛选已愈发常见。本文提出一种高效简洁的基于态函数的校正（SFC）算法，该算法利用自由能的态函数特性，无需进行环识别，从而消除了计算瓶颈，其计算复杂度为O(P×N)，其中P为微扰图中的边数，N为分子总数。与加权环闭合（WCC）等基于图的方法相比，SFC在图规模扩大时仍能维持稳定的计算效率，可高效处理多达50000个乃至更多分子的大型微扰网络，适用于高通量FEP-RBFE应用场景。此外，SFC还融入了不确定性感知加权机制，可进一步优化校正性能。上述优势使SFC成为一种高效的RBFE校正方法，能够更好地支撑面向先导化合物优化的高通量FEP-RBFE计算。