MACHINE LEARNING-DRIVEN PREDICTION OF FLUORESCENT PROBE PROPERTIES: BRIDGING THE GAP BETWEEN PREDICTION AND EXPERIMENTATION

Abstract

The development of organic fluorescent materials needs quick and precise predictions of photophysical characteristics for techniques like high throughput virtual screen ing. However, there is a challenge caused by the constraints of quantum mechanical computations, experiments, and time. This thesis investigates the field of machine learning-assisted fluorescence probe design to answer this difficulty. The main part of this investigation is the utilization of a substantial database of optical properties of organic compounds that was collected from various scientific papers. One of the complicating factors of this database is the presence of missing data which stems from the collection from various sources, and this inconsistency is examined with the use of a range of imputation methods. Furthermore, the thesis aims to construct predic tive models that can forecast properties that are inherent to fluorescent compounds such as quantum yield, absorption and emission spectra, among others. This research aims to pave the way for a more efficient and targeted approach to fluorescent probe design.