Prediction of Maximum Absorption Wavelength Using Deep Neural Networks

Fluorescent molecules are important tools in biologicaldetection, and numerous efforts have been made to develop compoundsto meet the desired photophysical properties. For example, tuning thewavelength allows an appropriate penetration depth with minimalinterference from the autofluorescence/scattering for a better signal-to-noise contrast. However, there are limited guidelines to rationally designor computationally predict the optical properties fromfirst principles,and factors like the solvent effects will make it more complicated.Herein, we established a database (SMFluo1) of 1181 solvated small-moleculefluorophores covering the ultraviolet-visible-near-infraredabsorption window and developed new machine learning models basedon deep neural networks for accurately predicting photophysicalparameters. The optimal system was applied to 120 out-of-samplecompounds, and it exhibited remarkable accuracy with a mean relativeerror of 1.52%. In this new paradigm, a deep learning algorithm is promising to complement conventional theoretical andexperimental studies offluorophores and to greatly accelerate the discovery of new dyes. Due to its simplicity and efficiency, datafrom newly developedfluorophores can be easily supplemented to this system to further improve the accuracy across various dyefamilies.