Quantitative structure–property relationship study of spectral properties of green fluorescent protein with support vector machine

Illuminating the origins of spectral properties of green fluorescent proteins via proteochemometric and molecular modeling

“…Before comparing model performances afforded by previously investigated QSPR models with the PCM models presented herein, it is pertinent to consider the methodological differences imposed by both. In light of conventional QSPR modeling, PCM modeling represents a leap forward for structure–activity/property relationship investigations owing to its ability to simultaneously incorporate descriptive information from several proteins and several ligands as well as its inherent interpretability in which the significance of descriptors can be elucidated.…”

“…The aforementioned investigations of spectral properties of GFP make use of sophisticated and time‐consuming approaches that may not be practically feasible. Nantasenamat et al proposed for the first time the construction of quantitative structure–property relationship (QSPR) models for predicting the spectral properties (i.e., excitation and emission maximas) of GFP using machine learning approaches. Such QSPR models represent a simple and rapid approach for discerning correlations between the structures and their investigated properties .…”

Molecular excited states through a machine learning lens

“…83 Descriptors inspired by drug-design studies can also include computationally affordable ground-state QM properties. 47,48 To ensure greater transferability and accuracy, ML models can be combined with QM methods in different ways (Box 3). Any such combination using QM method would obviously result in increased computational cost compared to pure ML approaches, which may be an obstacle when a large number of predictions are necessary, for example, in high-throughput screening.…”

“…178,179 For example, ML trained on excitation energies of fluorescent molecules can be used not only for emission spectrum simulation but also for the design of materials that emit at a required wavelength for application as LED or a fluorescent label in cell imaging. [47][48][49] More often than not, the compounds that are intended to be used in complex photodevices are selected based on macroscopic properties such as power conversion efficiency (PCE) or fluorescence quantum yield that ultimately depend on atomic-scale molecular excited-state properties. Calculating these properties with QM is often a formidable task and the beauty of ML is that it can be used for learning any of them.…”

High-throughput virtual screening for organic electronics: a comparative study of alternative strategies

“… 333 ML has also been used to predict the emission wavelength of multiple fluorescent organic molecules, using steric, hydrophobic and electronic properties. 334,335 Other examples of the prediction of excited state properties through machine learning and a more general view on its applicability can be found in ref. 336 and 337 .…”