A machine learning approach for the discovery of new chemical receptors using a microfluidic platform

具有高亲和力和选择性与特定分析物结合的分子受体受到广泛关注和重视。例如，它们已被用作化学传感器、用于细胞成像的光学探针，以及用于去除污染物的功能化固体材料的开发。然而，该领域最大的挑战之一是从头开始设计对给定分析物具有非常高特异性和亲和力的受体。通常，这要么是偶然实现的，要么是没有达到某些应用程序所需的水平。主要问题在于，受体-分析物结合事件对受体的几何形状、大小、电荷和结构中存在的化学基团高度敏感。这些参数的优化往往使受体的合成变得困难甚至不可行。为了克服这些问题，在本项目中，我们建议将用于高通量分子受体合成和测试的微流控平台与机器学习方法相结合，以用于新受体的设计。该项目最初将侧重于利用上述方法开发受体，以结合对环境有益的阴离子物种（例如砷酸盐、铬酸盐和磷酸盐的有毒衍生物）。然而，我们希望我们的结合微流体-机器学习的方法是通用的分子受体的发展为广泛的不同的分析物。

Molecular receptors that can bind to specific analytes with high affinity and selectivity are of widespread interest and importance. For example, they have been used as chemical sensors, optical probes for cellular imaging as well as in the development of functionalized solid materials for the removal of pollutants. However, one of the biggest challenges in this area is the design of receptors, from scratch, with very high specificity and affinity for a given analyte. Often, this is either achieved serendipitously or not to the levels required for some applications. The main problem resides in the fact that the receptor-analyte binding event is highly sensitive to the geometry of the receptor, its size, charge and the chemical groups present in the structure. The optimization of these parameters often makes the synthesis of the receptors difficult or even not feasible. To overcome these problems, in this project we propose to combine a microfluidic platform for the high-throughput synthesis and testing of molecular receptors, with machine learning approaches to feed into the design of the new receptors. The project will initially focus on using the above-mentioned approaches to develop receptors to bind anionic species of environmental interest (e.g. arsenate, chromate and toxic derivatives of phosphates). However, we expect our combined microfluidics-machine learning approach to be generic for the development of molecular receptors for a wide range of different analytes.