機器學習方法預測 數千種有機螢光團的放射波長

Abstract

在過去70年中，螢光分子已廣泛應用於各種領域，如螢光紡織品，螢光油墨和螢光塑料產品。 有機螢光顏料的應用也在螢光檢測，生物探針和標記方面。 在這項研究中，我們導入了超過一萬個有機螢光分子進行分析，並使用分子結構文件生成分子描述符。 我們還應用聚類方法，以更好地了解這些各種有機螢光分子並幫助建模。 我們希望為螢光分子的選擇和設計提供廣泛而有效的模型，並促進螢光材料的發展。在我們的信息方法處理之後，我們的模型中留下的一些描述符最初是為了描述環和多個鍵屬性而創建的。所選出的描述符與我們的化學直覺相關，且解釋性比重較高的描述符多為對共軛性質的描述，與以往化學家對螢光分子結構經驗相符。For over past 70 year fluorescent molecules have been used in a wide range of applications, such as Fluorescent textiles, Fluorescent ink, and Fluorescent plastic products. The applications of organic fluorescent pigments are also in the fluorescent inspection, biological probes and labeling aspect. In this research, we imported over ten thousands organic fluorescent molecules for analysis and use the molecular structure files to generate molecular descriptors. We apply the clustering method for the better insight in these various organic fluorescent molecules and aiding the modeling. We expect to bulid a broad and valid model for the selection and design of fluorescent molecules and spur the development of fluorescent materials. After our informatic methods process, some descriptors left in our model are originally created to describe the rings and multiple bond properties. The selected descriptors are closely related to our chemical instinct.