SPEAKER: Silvia Crivelli; Computational Chemistry, Materials & Climate, Lawrence Berkeley National Laboratory
TITLE: “Machine learning approaches for protein classification and protein-ligand binding prediction”
ABSTRACT: The combined use of high-performance computing, modeling and simulation, the massive amounts of data they generate in addition to data available from scientific instruments, and the data analytics successes reported by industry have fueled the interest in using machine learning for science and engineering. In particular, the predictive ability of machine learning to find patterns, signals, or structure that may be hidden within data sets (massive or not) is key to advances in biology, medicine and health care. However, machine learning methods developed for the industry cannot be straightforwardly applied to scientific applications and a great deal of work needs to be done to address issues associated with appropriate data representation, methodology, interpretability of results, and uncertainty quantification, among others. In this talk I will describe our year-long exploration for an efficient, accurate, and interpretable method for protein classification and protein-ligand binding prediction. I will also describe how we are applying deep learning technologies to improve health care outcomes by analyzing electronic health records from thousands of patients. I will provide an example aimed at improving the use of resources by predicting the likelihood that a patient suffering from a disease of despair and who has been hospitalized will be readmitted to the hospital within 30 days. Diseases of despair such as drug abuse, alcoholism, and suicide are contributing to declining US life expectancy.
DATE: Thursday, December 6th, 4:10pm
LOCATION: MSB 1147, Colloquium Room
REFRESHMENTS: 3:30pm MSB 4110 (4th floor lounge)
STA 290 Seminar List: https://statistics.ucdavis.edu/seminars