Exploring Deep Learning for Metalloporphyrins: Databases, Molecular Representations, and Model Architectures

Abstract

Metalloporphyrins have been studied as biomimetic catalysts for more than 120 years and have accumulated a large amount of data, which provides a solid foundation for deep learning to discover chemical trends and structure–function relationships. In this study, key components of deep learning of metalloporphyrins, including databases, molecular representations, and model architectures, were systematically investigated. A protocol to construct canonical SMILES for metalloporphyrins was proposed, which was then used to represent the two-dimensional structures of over 10,000 metalloporphyrins in an existing computational database. Subsequently, several state-of-the-art chemical deep learning models, including graph neural network-based models and natural language processing-based models, were employed to predict the energy gaps of metalloporphyrins. Two models showed satisfactory predictive performance (R2 0.94) with canonical SMILES as the only source of structural information. In addition, an unsupervised visualization algorithm was used to interpret the molecular features learned by the deep learning models.

Deep transfer learning for predicting frontier orbital energies of organic materials using small data and its application to porphyrin photocatalysts

“…62 For a more detailed description of the PBDD, see Table S1 in the ESI †. With the framework we developed previously, 33 the porphyrins dye structures in PBDD were converted to canonical SMILES. In the original PBDD, each porphyrin dye was represented by a row of short names for its side groups ( e.g.…”

“…2 Conversion of structural information from Porphyrin-based dyes database into canonical SMILES using the framework we developed. 33 The framework converts the short name of metalloporphyrin substituents and backbone (left) to SMILES fragments and assembles them to the full SMILES (bottom right) of the complexed compound.…”

“…64 Our group further refined the model for multitask property prediction of molecular complexes 42 and E_gap prediction of porphyrin dyes. 33 In this study, we developed Por-phyBERT by first pre-training the BERT-based model using the MpP structures of PBDD in an unsupervised learning manner (i.e. no property data,) and then fine-tuning the model with the MpPD containing both the structure and frontier orbital energy data.…”

“…31,32 Our group recently developed a framework to convert the intrinsically non-generic molecular representation of the PBDD porphyrins to SMILES, a universal molecular representation for chemical machine learning. 33 With the help of this framework, the PBDD was used to train multiple deep neural network models to predict the energy gap of porphyrin dyes. However, we must emphasize that the porphyrin dyes in the PBDD database are used as photosensitizers, 31,32 and their structural design is significantly different from that of photocatalysts (i.e., porphyrins used as photosensitizers have a unique anchor group that can be attached to semiconductors).…”

Semantic Hybrid Signal Temporal Logic Learning-Based Data-Driven Anomaly Detection in the Textile Process

“…With the rapid evolution of computational power, more machine learning algorithms have demonstrated powerful roles in the practical application of anomaly detection in the textile industry [4]. According to the detected abnormality, the operator can take further measures to avoid serious failures [5,6].…”