PolyNC: a natural and chemical language model for the prediction of unified polymer properties

Qiu, Haoke; Liu, Lunyang; Qiu, Xuepeng; Dai, Xuemin; Ji, Xiangling; Sun, Zhao-Yan

doi:10.1039/d3sc05079c

Cited by 3 publications

(4 citation statements)

References 61 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The DSC curves can be found in the Supporting Information S6.4. The DSC results for PI-1b and PI-2b are obtained from our previous work . These results demonstrated the consistency between our model’s predicted values and the ground truth values.…”

Section: Resultssupporting

confidence: 81%

Heat-Resistant Polymer Discovery by Utilizing Interpretable Graph Neural Network with Small Data

Qiu,

Wang,

Qiu

et al. 2024

Macromolecules

Self Cite

View full text Add to dashboard Cite

Polymers with exceptional heat resistance are critically valuable in numerous domains, particularly as essential components of flexible organic light-emitting diodes. Among these, polyimides (PIs) demonstrate significant potential as substrate candidates for these next-generation flexible displays due to their robustness. However, traditional Edisonian approaches struggle to navigate the vast chemical space of PIs and also pose challenges of small data, which constrains the learnable chemical space for machine learning (ML). In this study, we propose a chemical-knowledge-based strategy to facilitate the design of PIs with high glass transition temperature (T g ) utilizing an atom-wise graph neural network and small data. Inspired by chemical intuition, our strategy leverages the available data on the same property (i.e., T g ) from other polymers, which is beneficial for expanding the chemical space used for ML. The trained ML model achieves an impressive performance in predicting T g of polymers. We have also investigated the impact of the chemical space encompassed by the data sets on the performance of ML models. Through interpretability analysis, it has been demonstrated that our ML model has learned more accurate chemical knowledge. Utilizing the ML model, 89 PIs were rapidly discovered from over 10 6 candidates, with experimental validation confirmed their exceptional heat resistance of the most promising PIs, which have been found to possess a T g exceeding 405 °C and even 450 °C. These results, along with the trained ML model, have the potential to accelerate the discovery of polymer substrate materials for next-generation flexible display devices.

show abstract

Section: Resultssupporting

confidence: 81%

Heat-Resistant Polymer Discovery by Utilizing Interpretable Graph Neural Network with Small Data

Qiu,

Wang,

Qiu

et al. 2024

Macromolecules

Self Cite

View full text Add to dashboard Cite

show abstract

“…Several established chemical language models possess the ability to comprehend chemical rules and generate novel chemical substances. Recently, there have been some advancements in polymer chemical language models, such as TransPolymer, PolyBERT, and PolyNC, which aim to address the data and modeling challenges of polymers and predict their various properties in a vast structural space. Polymer large models trained through data parallelism can effectively integrate and utilize the data from different sources and modalities.…”

Section: Discussionmentioning

confidence: 99%

“…Drawing inspiration from natural language processing methods, the polymer informatics field is leveraging chemical language to establish large models for polymer structure generation and property prediction. ,− A large model refers to a model with extensive and intricate parameters, such as the large model of Chat Generative Pretrained Transformer (ChatGPT). Several established chemical language models possess the ability to comprehend chemical rules and generate novel chemical substances.…”

Section: Discussionmentioning

confidence: 99%

“…Drawing inspiration from natural language processing methods, the polymer informatics field is leveraging chemical language to establish large models for polymer structure generation and property prediction. 53 PolyBERT, 54 and PolyNC, 56 which aim to address the data and modeling challenges of polymers and predict their various properties in a vast structural space.…”

Section: Establishing Polymer Large Models Based On Chemical Languagementioning

confidence: 99%

See 1 more Smart Citation

Machine Learning-Assisted Design of Advanced Polymeric Materials

Gao,

Lin,

Wang

et al. 2024

Acc. Mater. Res.

View full text Add to dashboard Cite

Conspectus Polymeric material research is encountering a new paradigm driven by machine learning (ML) and big data. The ML-assisted design has proven to be a successful approach for designing novel high-performance polymeric materials. This goal is mainly achieved through the following procedure: structure representation and database construction, establishment of a ML-based property prediction model, virtual design and high-throughput screening. The key to this approach lies in training ML models that delineate structure–property relationships based on available polymer data (e.g., structure, component, and property data), enabling the screening of promising polymers that satisfy the targeted property requirements. However, the relative scarcity of high-quality polymer data and the complex polymeric multiscale structure–property relationships pose challenges for this ML-assisted design method, such as data and modeling challenges. In this Account, we summarize the state-of-the-art advancements concerning the ML-assisted design of polymeric materials. Regarding structure representation and database construction, the digital representations of polymers are the predominant methods in cheminformatics along with some newly developed methods that integrate the polymeric multiscale structure characteristics. When establishing a ML-based property prediction model, the key is choosing and optimizing ML models to attain high-precision predictions across a vast chemical structure space. Advanced ML algorithms, such as transfer learning and multitask learning, have been utilized to address the data and modeling challenges. During the ML-assisted screening process, by defining and combining polymer genes, virtual polymer candidates are generated, and subsequently, their properties are predicted and high-throughput screened using ML property prediction models. Finally, the promising polymers identified through this approach are verified by computer simulations and experiments. We provide an overview of our recent efforts toward developing ML-assisted design approaches for discovering advanced polymeric materials and emphasize the intricate nature of polymer structural design. To well describe the multiscale structures of polymers, new structure representation methods, such as polymer fingerprint and cross-linking descriptors, were developed. Moreover, a multifidelity learning method was proposed to leverage the multisource isomerous polymer data from experiments and simulations. Additionally, graph neural networks and Bayesian optimization methods have been developed and applied for predicting polymer properties as well as designing polymer structures and compositions. Finally, we identify the current challenges and point out the development directions in this emerging field. It is highly desirable to establish new structure representation and advanced ML modeling methods for polymeric materials, particularly when constructing polymer large models based on chemical language. Through this Account, we seek to stimulate further ...

show abstract

Tutorial: AI-assisted exploration and active design of polymers with high intrinsic thermal conductivity

Huang,

2024

Journal of Applied Physics

View full text Add to dashboard Cite

Designing polymers with high intrinsic thermal conductivity (TC) is critically important for the thermal management of organic electronics and photonics. However, this is a challenging task owing to the diversity of the chemical space and the barriers to advanced synthetic experiments/characterization techniques for polymers. In this Tutorial, the fundamentals and implementation of combining classical molecular dynamics simulation and machine learning (ML) for the development of polymers with high TC are comprehensively introduced. We begin by describing the core components of a universal ML framework, involving polymer data sets, property calculators, feature engineering, and informatics algorithms. Then, the process of constructing interpretable regression algorithms for TC prediction is introduced, aiming to extract the underlying relationships between microstructures and TCs for polymers. We also explore the design of sequence-ordered polymers with high TC using lightweight and mainstream active learning algorithms. Lastly, we conclude by addressing the current limitations and suggesting potential avenues for future research on this topic.

show abstract

PolyNC: a natural and chemical language model for the prediction of unified polymer properties

Abstract: PolyNC directly infers properties based on human prompts and polymer structures, enabling an end-to-end learning that encourages the model to autonomously acquire fundamental polymer knowledge, in a multi-task, multi-type unified model manner.

Cited by 3 publications

References 61 publications

Heat-Resistant Polymer Discovery by Utilizing Interpretable Graph Neural Network with Small Data

Heat-Resistant Polymer Discovery by Utilizing Interpretable Graph Neural Network with Small Data

Machine Learning-Assisted Design of Advanced Polymeric Materials

Tutorial: AI-assisted exploration and active design of polymers with high intrinsic thermal conductivity

Contact Info

Product

Resources

About