Converting SMILES to Stacking Interaction Energies

Predicting
the strength of stacking interactions involving heterocycles
is vital for several fields, including structure-based drug design.
While quantum chemical computations can provide accurate stacking
interaction energies, these come at a steep computational cost. To
address this challenge, we recently developed quantitative predictive
models of stacking interactions between druglike heterocycles and
the aromatic amino acids Phe, Tyr, and Trp (DOI: 10.1021/jacs.9b00936). These models depend on heterocycle descriptors derived from electrostatic
potentials (ESPs) computed using density functional theory and provide
accurate stacking interactions without the need for expensive computations
on stacked dimers. Herein, we show that these ESP-based descriptors
can be reliably evaluated directly from the atom connectivity of the
heterocycle, providing a means of predicting both the descriptors
and the potential for a given heterocycle to engage in stacking interactions
without resorting to any quantum chemical computations. This enables
the rapid conversion of simple molecular representations (e.g., SMILES)
directly into accurate stacking interaction energies using a freely
available online tool, thereby providing a way to rank the stacking
abilities of large sets of heterocycles.

预测含杂环（heterocycles）的堆积相互作用强度，对于包括基于结构的药物设计在内的多个研究领域至关重要。尽管量子化学计算可获得准确的堆积相互作用能，但其计算成本高昂。为解决这一难题，我们近期开发了类药杂环与芳香族氨基酸苯丙氨酸（Phe）、酪氨酸（Tyr）及色氨酸（Trp）之间堆积相互作用的定量预测模型（DOI: 10.1021/jacs.9b00936）。该类模型基于通过密度泛函理论计算得到的静电势（electrostatic potentials, ESPs）衍生的杂环描述符构建，无需对堆积二聚体开展高成本计算即可得到准确的堆积相互作用结果。本文研究表明，可直接通过杂环的原子连接性可靠计算得到上述基于ESPs的描述符，从而无需借助任何量子化学计算，即可同时预测该描述符以及特定杂环参与堆积相互作用的潜力。借助一款免费可用的在线工具，该方法可将简单的分子表征（如简化分子线性输入规范（SMILES））快速直接转换为准确的堆积相互作用能，进而实现对大量杂环堆积能力的排序。