D3SC: EAGER: Data-driven development of fluorescent sensors for bio-imaging

Washington State University, Pullman WA

Investigators

Abstract

Chemical information is growing dramatically, fueled by the massive amount of data generated by researchers in a variety of fields. Conventional approaches require that scientists sift through data from many sources to try to develop a comprehensive picture of the field when making a new chemical compound or studying a chemical processes. This conventional approach is slow and difficult as researchers can miss important trends. With the support from the Chemical Measurement and Imaging (CMI) Program in the Division of Chemistry and the Cyberinfrastructure for Emerging Science and Engineering Research (CESER) Program in the Office of Advanced Cyberinfrastructure, Professor Xian at Washington State University (WSU) and Professor Ji are teaching machines to collect massive data from literature, analyze them, and come up with new design principles to make new sensors. The project is in response to the Data-Driven Discovery Science in Chemistry Dear Colleague Letter (D3SC-DCL). The team is using data mining and computer learning to develop a new generation of sensors for the detection of hydrogen sulfide (H2S). Hydrogen sulfide is an important signaling molecule that is associated with biological processes such as high blood pressure, atherosclerosis, coronary heart diseases. The coupling of computer science techniques with chemical problems enables Professors Xian and Ji and their students to predict the most promising sensor design without making and testing a large number of sensors empirically. The research group "trains their computer" by testing the predicted sensors designs and providing feedback for the next iteration; this is machine learning. If successful, the machine learning methods may be expanded to sensor development for the detection of many different compounds, especially those that are important in biological and chemical processes (such as biological warfare and pharmaceutical development). The research provides unique training opportunities for undergraduate and graduate students by providing rich experiences in both chemistry and data science. These activities build the workforce of non-traditional chemistry trainees to meet data-driven research and development needs in industry and academia. Professor Xian also actively works with undergraduate students from underrepresented minority groups by participating the Pacific Northwest Louis Stokes Alliance for Minority Participation (PNW-LSAMP) Program at WSU. Professors Xian and Ji are building a database of H2S sensors with searchable parameters. They are carrying out data-driven optimization of the sensors based on advanced machine learning and data mining techniques. A combined data-driven discovery framework of unsupervised learning and supervised multi-task learning is developed to predict the important properties of the sensors. This approach is used to identify the most suitable fluorophores and H2S-reaction sites for the design of optimal sensors, which can then be synthesized and validated. These studies may advance our understanding of machine learning and data mining as well as chemical predication and sensor development. This interdisciplinary project provides a unique platform to attract students to chemistry and train them at the interface of chemistry, data science, and computation science. Broadening participation efforts include students from underrepresented minority groups through the Pacific Northwest Louis Stokes Alliance for Minority Participation (PNW-LSAMP) Program at WSU.