/tag/data-science
Deep Learning for Glaucoma Patient Risk Stratification
Glaucoma is a chronic eye disease that can lead to irreversible vision loss if not properly managed. Effectively stratifying patient risk is a major clinical challenge, as clinicians need tools to identify patients at high risk of progression from sparse and irregularly-sampled electronic health records. Traditional forecasting methods struggle with the irregular time-series data typical of EHRs, where features vary with visits and occur at different intervals. Towards developing these critical risk identification tools, DAC consultants collaborated with Dr. Heman Shakeri’s lab from the School of Data Science as well as with advice from clinicians to develop a novel deep kernel learning architecture that leverages a Gaussian Process backend with a transformer-based feature extractor to model glaucoma patient trajectories from multimodal EHR data.
In-Silico NCAA-Containing Peptide Design
Bacteria are an important type of human pathogen that can cause life-threatening infections. Increasingly, these microorganisms can survive the effects of antibiotics previously used to kill them. As bacteria become resistant to multiple kinds of antibiotics, the diseases they cause become ever more difficult to cure. Accordingly, infections caused by ‘multidrug-resistant’ (MDR) pathogens are associated with frequent treatment failures, high hospitalization costs, and substantial mortality. New therapeutics are needed to treat infections caused by MDR bacteria. Towards developing these critical countermeasures, our group has discovered a unique peptide that efficiently kills many of the most challenging antibiotic-resistant pathogens and also demonstrates therapeutic efficacy in pre-clinical animal models of bacterial infection.
Political Sentiment Analysis
The nature of political communication has been fundamentally altered by the emergence of social media. In earlier eras, social scientists, journalists, and citizens could focus on static statements by politicians and candidates in order to understand the nature of political discourse. Social scientists studying political communication would design surveys and focus groups to understand which messages were received by citizens, and with what effect. Today, as news moves to digital platforms and as political figures increasingly rely on social media, political communication is fundamentally dynamic. Studying patterns of communication among politicians, their supporters, and their critics requires scholarly focus on the content, sentiment, and framing of posts on various social media platforms.
