Calculating free energies of binding (ΔGbind) between ligands and their target protein is of major interest to drug discovery and safety, yet it is still associated with several challenges and difficulties. Linear interaction energy (LIE) is an efficient in silico method for ΔGbind computation. LIE models can be trained and used to directly calculate binding affinities from interaction energies involving ligands in the bound and unbound states only, and LIE can be combined with statistical weighting to calculate ΔGbind for flexible proteins that may bind their ligands in multiple orientations. Here, we investigate if LIE predictions can be effectively improved by explicitly including the entropy of (de)solvation into our free-energy calculations. For that purpose, we combine LIE calculations for the protein-ligand-bound state with explicit free-energy perturbation to rigorously compute the unbound ligand's solvation free energy. We show that for 28 Cytochrome P450 2A6 (CYP2A6) ligands, coupling LIE with alchemical solvation free-energy calculation helps to improve obtained correlation between computed and reference (experimental) binding data.

中文翻译：

---

结合线性相互作用能和炼金溶剂化自由能方法进行蛋白质结合亲和力计算。

计算配体与其靶蛋白之间的结合自由能（ΔGbind）是药物发现和安全性的主要关注点，但它仍然面临一些挑战和困难。线性相互作用能（LIE）是用于ΔGbind计算的一种有效的计算机方法。可以训练LIE模型并将其用于仅从涉及结合态和未结合态的配体的相互作用能直接计算结合亲和力，并且LIE可以与统计权重结合以计算可在多个方向结合其配体的柔性蛋白质的ΔGbind。在这里，我们调查是否可以通过将（去）溶剂化的熵明确包含在我们的自由能计算中来有效地改善LIE预测。为了这个目的，我们将蛋白质配体结合状态的LIE计算与显式自由能扰动结合起来，以严格计算未结合配体的溶剂化自由能。我们表明，对于28种细胞色素P450 2A6（CYP2A6）配体，将LIE与炼金溶剂化自由能计算耦合有助于改善计算和参考（实验）结合数据之间的相关性。