Biologically Informed Single-cell Data for Deep Learning

Single cell data has gained popularity for its cell level resolution and wide areas of applications such as cell-type annotation, cell trajectory analysis to understand development, disease mechanism and treatment effects. Large consortiums are built to provide author-provided processed data from many studies, with some metadata curation added on top to enable expedited analysis. However, studies using different processing pipelines, references and batch correction methods add confounding factors in the downstream usecases which require integrating multiple single cell datasets. In addition, biological variation and knowledge bases are often not captured to enrich the analysis with biological context. In this webinar we will showcase some insights from published literature and internal work that support the re-analysis of single cell data to have consistent, biologically informed high quality AI-ready datasets.