Network and Machine Learning Methods for Studying Evolution of Biological Networks Applied to Predictive Medicine

Date/Time
Date(s) - 02/21/2022
3:00 pm - 4:00 pm

Location
Virtual via Zoom & projected in Communicore, C1-004

Sanjukta Krishnagopal, Ph.D., Postdoctoral Researcher, Gatsby Computational Neuroscience Unit, University College London

Join Zoom Meeting:
Launch Zoom Meeting

Networks are a natural way to model time-varying interactions between elements of biological systems such as neurons in the brain or gene-expression patterns in diseases. First, I introduce a novel network-based algorithm for identifying and predicting subtypes in diseases such as Parkinson’s and Stroke that are characterized by time-varying interactions between various factors/symptoms. Specifically, I develop a trajectory-based algorithm that predicts Parkinson’s disease subtype with ~70% accuracy five years in advance, demonstrating the usefulness of the method in personalized medicine. Our subtypes are clinically relevant, supported by clinical knowledge, and are generalizable to various types of data – which I demonstrate through studying genotypic markers unique to each subtype. The importance of time of medication administration is explored through graph neural networks. In the second part of the talk, I study network interactions in the brain, and introduce a novel biologically plausible machine learning model, bridging the fields of artificial and cognitive intelligence. This model, called Dendritic Gated Network (DGN), maps directly onto the Cerebellum, and is trained on learning a sequence of changing tasks. We find that DGNs are significantly more efficient than conventional artificial neural networks and possess several desirable properties like having local learning and remembering old tasks. Through our model, we make predictions about interneuron activity in the cerebellum, validated through experiments.

Biography: Sanjukta’s research lies at the interface of computational biology, network science, machine learning and complex systems, with the goal of addressing real world problems in biological systems, from molecular to connectome-wide scales. She received her PhD from the University of Maryland in Physics where she was also a fellow of the CoMBiNe (Computational and Mathematics in Biological Networks) fellowship. She is currently a postdoc at the Gatsby Computational Neuroscience Unit at University College London. She develops interdisciplinary computational and mathematical tools, often motivated by experimental evidence and typically involving various types of data analysis. She is interested in developing methods for analyzing, subtyping, and treating a variety of diseases such as cancer and neurodegenerative disorders, as well as studying the role of gene expression in them. She has lived in Mumbai, Berlin, Washington DC and Berlin, and enjoys dancing, diving and hiking in her spare time.