AI-ZYMES

AI-ZYMES, the website of the Combination of Nanazyme and Artificial Intelligence is an online platform focusing on the research and discussion of the cross-application of nanozyme technology and artificial intelligence in the field of chemistry and computer. This website brings together the latest research results, technological innovations and future trends from experts and researchers in different fields. By providing rich article data and functional predictions, this website aims to promote interdisciplinary communication and cooperation, and promote the development and application of nanozymes and artificial intelligence technologies.

Nanozymes is a series of nanomaterials with intrinsic enzyme-like properties. By using artificial intelligence algorithm and large language model, the properties and synthesis paths of existing nanozymes were screened, and a comprehensive and complete database involving nanozymes in existing published articles was established, which provided a research basis for data calculation, experimental design and property prediction of nanozymes. By collecting data from existing articles, a user-friendly and scalable database of more than 400 existing inorganic nanozymes was developed. Based on machine learning method, artificial intelligence algorithm is applied to the intelligent design of catalytic materials and nanozymes, text analysis of the synthesis path of nanozymes is carried out, and guidance development of new nanozymes with ideal properties is promoted. Main application fields: Chemistry, computer. Service scenario: chemical nanozymes research institutions, nanozymes beginners understanding and learning, laboratory design experiment use prediction, etc.

Main Advantage

  • 1. ChatGPT and human -guiding data collection: ChatGPT 4.0 was utilized for initial data extraction, by optimizing prompt engineering, the ChatGPT’s text mining accuracy reached up to 67.55%. Considering the unavoidable “Not specified” values caused by ChatGPT, we used group cooperation for the second-round manual data collection, addressing instances where data were presented in figures rather than the text section. Reliable data, by ChatGPT and human -guiding data collection, A total of 1085 pieces of data were collected, and our team ensured the strict reliability and referrability of the data.
  • 2. Compared with the accuracy of previously reported RFR and DNN models, Gradient Boosting Regression (GBR) model demonstrated superior prediction performance for nanozyme catalytic activities (R2=0.6476 for Km and R2= 0.95 for K cat). Website provides accurate predictions! The accuracy of this study is the highest among the publicly available databases.
  • 3. A ChatGPT-based nanozyme copilot was developed for guiding the synthesis of nanozymes. Retrieval augmented generation (Ragas) was firstly introduced to verify the accuracy of the nanozyme copilot’s response, which reached more than 90%.