The goal of our research is to develop machine learning methods for real-world settings and use them to enable new discoveries in biology and medicine. We work towards inventing machine learning methods that are applicable to complex and heterogeneous real-world data and have the ability to generalize to novel scenarios that have never been encountered during model training. We apply our methods to solve cutting-edge problems in biomedical research.
Advancing machine learning research
Our machine learning methods are inspired by challenges that arise when we deal with real-world data in which conventional machine learning assumptions are not satisfied. In real-world applications such as biomedicine, datasets often originate from different experiments. Thus, train and test data do not need to come from the same distribution or even from a single learning task. In addition, new categories that have never been encountered during training can appear during model deployment. We design methods that can generalize to new learning tasks with minimal supervision, e.g., only a few labeled training examples, or even fully unsupervised. Our research directions can be broadly categorized in the areas of meta-learning and few-shot learning, unsupervised learning, semi-supervised learning, open-world and open-set learning, domain adaptation and domain generalization, graph representation learning, explainable AI.
Example publications:
- The Pursuit of Human Labeling: A New Perspective on Unsupervised Learning
- Towards Universal Cell Embeddings: Integrating Single-cell RNA-seq Datasets across Species with SATURN
- Open-World Semi-Supervised Learning
- Annotation of Spatially Resolved Single-cell Data with STELLAR
- Concept Learners for Few-Shot Learning
- MARS: Discovering Novel Cell Types across Heterogeneous Single-Cell Experiments
- Selective Sampling-Based Scalable Sparse Subspace Clustering
- Multi-view Low-rank Sparse Subspace Clustering
- l0-Motivated Low-Rank Sparse Subspace Clustering
Advancing biomedical research
We use our methods to unravel new biological insights and work on transformative applications of machine learning in biomedicine. We are especially interested in using our methods for the analysis and understanding of high throughput single-cell data which provides unprecedented resolution to examine cell heterogeneity in tissues and organs. Our methods have been used to annotate global cell atlas consortia efforts aiming at creating comprehensive reference maps of all cell types such as HuBMAP and Fly Cell Atlas. We intensively collaborate with biologists, neuroscientists and medical researchers on cutting-edge problems across fields and disciplines.
Example publications:
- Aging Fly Cell Atlas Identifies Exhaustive Aging Features at Cellular Resolution
- High Resolution Single Cell Maps Reveals Distinct Cell Organization and Function Across Different Regions of the Human Intestine
- Fly Cell Atlas: A Single-Cell Transcriptomic Atlas of the Adult Fruit Fly
- Single-Cell Transcriptomes of Developing and Adult Olfactory Receptor Neurons in Drosophila
- Temporal evolution of Single-Cell Transcriptomes of Drosophila Olfactory Projection Neurons
- The Landscape of Microbial Phenotypic Traits and Associated Genes