CCB Seminar: Deep Representations for Learning for Problems in Biology

Dr. Smita Krishnaswamy, Assistant Professor, School of Medicine, Yale
Abstract:
High throughput, high dimensional data is now ubiquitous in biomedical science, bolstered by advances in measurement technologies such as single cell RNA-sequencing, single cell ATAC-sequencing. Such data can give us an unprecedented view into cellular systems and patient response. However, current approaches are unable to derive maximal predictive insight from this data, as many initial efforts were largely focused on denoising, batch-normalization and other pre-processing tasks. We propose to tackle this by forming multiscale representations of the data based on data geometry and deep learning systems. In particular, I will present an embedding called Multiscale PHATE that allows for exploration of structure and meaningful, predictive abstractions of the data in a scalable fashion. We show results of MultiScale PHATE on COVID-19 facs and viral genomic sequence data. When combined with our MELD method for comparing two or more single-datasets, we are able to discover small populations of cells that are differentially expressed in patients with different outcomes of disease. Further, we will show that after dimensionality reduction we can learn the continuous dynamics of data from static snapshot measurements using our neural ODE framework called TrajectoryNET. Together these tools can start to help us gain overall insight into the structures, dynamics, and predictive features of the data.
Bio:
Smita Krishnaswamy is an Assistant professor in Genetics and Computer Science. She is affiliated with the applied math program, computational biology program, Yale Center for Biomedical Data Science and Yale Cancer Center. Her lab works on the development of machine learning techniques to analyze high dimensional high throughput biomedical data. Her focus is on unsupervised machine learning methods, specifically manifold learning and deep learning techniques for detecting structure and patterns in data. She has developed algorithms for non-linear dimensionality reduction and visualization, learning data geometry, denoising, imputation, inference of multi-granular structure, and inference of feature networks from big data. Her group has applied these techniques to many data types such as single cell RNA-sequencing, mass cytometry, electronic health record, and connectomic data from a variety of systems. Specific application areas include immunology, immunotherapy, cancer, neuroscience, developmental biology and health outcomes. Smita has a Ph.D. in Computer Science and Engineering from the University of Michigan. Following her Ph.D., Smita worked as a researcher at IBM’s TJ Watson research center in the systems division, and then completed her postdoctoral training in Systems Biology from Columbia University.