Integrating Single-Cell Data with Substantial Batch Effects - Improving cVAE Regularization

Explore a comprehensive lecture on integrating single-cell data with substantial batch effects. Learn about the challenges and solutions in combining multiple datasets for new insights, including public perturbation screens and comparisons of preclinical models. Discover novel approaches to address batch effects using conditional variational autoencoders (cVAEs) with advanced regularization techniques. Understand the benefits of VampPrior and latent cycle-consistency loss in preserving biological information while enhancing batch correction. Gain insights into the newly proposed model combining these techniques and its implementation in the scvi-tools package as sysVI. Follow the speaker's journey through background information, single-cell integration methods, batch effect removal strategies, and techniques for preserving biological information. Conclude with a discussion on future applications and participate in a Q&A session to deepen your understanding of this cutting-edge approach to single-cell data analysis.