Data-driven design of polymers for possible use for fluorescent applications

The exploration and designing of polymers for fluorescent applications is a subject of significant interest. This study presents a novel technique for the designing of fluorescent polymers, machine learning (ML) analysis using molecular descriptors is used. Through statistical methods, the most effective molecular descriptors (features) are identified. For the prediction of photoluminescence quantum yield (PLQY), the K neighbors regressor and extra trees regressors are employed by means of these optimal descriptors. To generate a diverse set of polymers, the Breaking Retro-synthetically Interesting Chemical Substructures (BRICS) method is employed and 10,000 new polymers are generated. The selected thirty polymers are afterwards built based upon their predicted PLQY values. To acquire a comprehension about the chemicals character, the chemical similarity analysis is carried out. Further, the clustering and heatmap techniques are utilized for this purpose. This research is expected to provide valuable guidance for experimental chemists in the synthesis of efficient fluorescent polymers.