Bio
Prof. Tijana Milenkovic, a Frank M. Freimann Collegiate Professor of Engineering, has been a faculty member at the University of Notre Dame since 2010, after earning her Ph.D. degree in Computer Science from the University of California Irvine in the same year. Prof. Milenkovic was promoted to an Associate Professor with tenure in 2016 and to a Professor in 2021. Milenkovic lab solves challenging problems in the fields of network science, graph algorithms, computational biology, scientific wellness, and social networks. Prof. Milenkovic won 2015 NSF CAREER, 2016 AFOSR YIP, and 2021 CRA-E Undergraduate Research Faculty Mentoring Awards, among others. She has published 53 journal papers (e.g., in Science, PNAS, or Bioinformatics) and 16 conference papers (e.g., in the top tier ISMB/ECCB computational biology conference). Prof. Milenkovic was elected to the Board of Directors of the International Society for Computational Biology (ISCB) in 2020 to represent the Society’s Communities of Special Interest (COSIs). She has been an active organizer of the flagship Network Biology (NetBio) COSI meeting at ISMB/ECCB since 2017, and of the Great Lakes Bioinformatics Conference since 2015. Prof. Milenkovic has been an Associate Editor of IEEE/ACM TCBB since 2014, of Scientific Reports since 2018, and of Frontiers in Bioinformatics -- Network Bioinformatics since 2020. Prof. Milenkovic is committed to increasing participation of women and diversity in Computer Science.
Talks
-
Abstract
Network science reveals a protein's role in aging and a person's risk of mental health problems
Keynote
02-09-2022 - 17:15-18:00
Networks (or graphs) are powerful models of complex systems in various domains, from biological cells to societies to the Internet. How to efficiently study these data, especially with increasing availability of dynamic (temporal) real-world networks? This talk will discuss our state-of-the-art computational approaches for network analysis, including those for studying dynamic networks, some of which are based on graphlets (subgraphs, Lego-like building blocks of complex networks). Also, this talk will demonstrate usefulness of our (dynamic) network analysis approaches in two tasks: studying the role of a protein in the aging process based on its position in a (dynamic) molecular network, and analyzing an individual's mental health based on their position in a (dynamic) social network, as follows.
First, incidence of many complex diseases, such as cancer, Alzheimer's disease, and even COVID-19 increases with age. Understanding the molecular mechanisms behind the aging process, including identification of human genes (i.e., their protein products) implicated in aging, is important for treating such aging-related diseases. However, wet lab experimental analyses of human aging are hard due to long human life span and ethical constraints. Computational identification (i.e., prediction) of aging-related genes via machine learning from human -omics data can fill in this gap. In this context, we integrated aging-specific gene expression data with context-unspecific protein-protein interaction (PPI) network data to infer a dynamic aging-specific PPI subnetwork. Then, we developed a machine learning model that when applied to the dynamic subnetwork can analyze how genes' PPIs change with age. So, our predictive model could guide the discovery of novel aging-related gene candidates for future wet lab validation.
Second, mental disorders such as depression and anxiety are public health issues. Early interventions can significantly reduce risk of developing mental disorders. Yet, most people do not seek treatments due to a lack of awareness of their disorders. A way to raise awareness is to develop computational approaches for predicting whether and when an individual will become at risk of a mental disorder. Innovative technologies such as wearable sensors can provide a wealth of data relevant to mental health. In this context, we leveraged rich longitudinal data from the recent NetHealth study containing individuals' social interaction data collected via smartphones, health-related behavioral data (physical activity and sleep duration) collected via Fitbit devices, and a variety of individuals' trait data (including mental health) collected via surveys. We modeled the NetHealth data as a dynamic network and developed a machine learning model for predicting one's likelihood of being depressed or anxious based on how their position in the dynamic network changes over time.
