ChromFormer: A transformer-based model for 3D genome structure prediction
- Henry Valeyre
- Pushpak Pati
- et al.
- NeurIPS 2022
Pushpak Pati is a Postdoctoral researcher at IBM Research Europe in Zurich in AI for single-cell research. He received his M.Sc. degree in Electrical Engineering specializing in computer vision and machine learning from ETH Zurich, Switzerland, in 2017. His Ph.D. research was carried out working jointly in the Computer Vision Lab at ETH Zurich and IBM Research Europe.
Pushpak's research focuses on modeling spatial tissue microenvironments in terms of biologically comprehensible elements across different biological stains, and understanding how it affects cancer examination, response to treatment, and biomarker identification. To achieve that, he combines deep learning and computer vision approaches to develop computational methods able to extract biologically meaningful patterns from large-scale, heterogeneous, and noisy tissue imaging data. His research also addresses well-known limitations in deep learning related to annotation scarcity, scalability to large image dimensions, and interpretability.
Pushpak has also been the main developer of HistoCartography, that facilitates the development of graph-based computational pathology pipelines, and has been contributing to ATHENA, that facilitates the visualization, processing, and analysis of (spatial) heterogeneity from spatial omics data.