Search

Deep learning has brought a dramatic development in molecular property prediction which is crucial in the field of drug discovery. Various methods such as fingerprint, SMILES, graphs have been proposed for representing molecules. Recently, unlabeled molecule data is used to improve performance for various pre-training methods. The main challenge of molecular properties predictions is designing a data representation and model that can show good performance for various datasets. However, performance deviation due to scarcity of dataset exists in constructing the model. We propose a new self-supervised method to learn the characteristics and structures of molecules by integrating existing methods. The key of our model is learning structures with matrix embedding and learning logics that can infer descriptors via QED prediction. As a result, our method improves the generalization of the data and achieves the best average performance by benchmarking downstream tasks. Moreover, we develop a web-based fine-tuning service to utilize our model on various tasks.

Recent impressive achievements on artificial intelligence and its technologies have expected to bring our daily life to the yet-experienced new world. Such technologies have also applied in the literature of materials science especially on data-driven materials research so that they could reduce the computing resources and alleviate redundant simulations. Nevertheless, since these are still in the immature stage, most of the datasets are private and have made according to their own standards and policies, therefore, they are hard to be merged as well as analyzed together. We have developed the Scientific Data Repository platform to store various and complicated data including materials data, which can analyze such data on the web. As the second step, we develop a machine learning based materials properties prediction tool enabling the fast discovery of advanced materials by using the general-purpose high-precise formation energy prediction module that performs MAE 0.066 within 10 s on the web.