Computational design of molecules for optimal performance is of interest in many fields, including chemical engineering. Often, however, these methods, in particular those based on rigorous mathematical optimization, do not explicitly take into consideration chemistry information, such as (but not limited to) synthesis feasibility. This opinion article discusses traditional and current approaches through examples from the literature where properties that depend on chemical transformations of the molecule are incorporated in the design process. Through these examples, the article highlights the importance of cheminformatics, graph theory, and machine learning in: (1) representation of the molecules, (2) reaction prediction and generation, and (3) property estimation. The article finally presents a vision of including information about chemical transformations in molecule design procedures, highlighting rigorous optimization and machine learning approaches such as generative modeling and reinforcement learning.

为获得最佳性能而对分子进行计算设计在许多领域都受到关注，包括化学工程。然而，这些方法，特别是那些基于严格数学优化的方法，通常没有明确考虑化学信息，例如（但不限于）合成可行性。这篇意见文章通过文献中的例子讨论了传统和当前的方法，其中依赖于分子化学转化的特性被纳入设计过程。通过这些例子，文章强调了化学信息学、图论和机器学习在以下方面的重要性：(1) 分子表示，(2) 反应预测和生成，以及 (3) 特性估计。