Multimodal Integration of scRNA-seq and ATAC-seq Data for Optimizing Latent Representations

This thesis investigates multimodal representation learning for high-dimensional, sparse biological data. The core task is to develop a machine learning model to jointly embed scRNA-seq (gene expression) and ATAC-seq (chromatin accessibility) data. The objective is to generate a unified latent representation that captures complex cross-modal dependencies, aiming to prove superior to single-modality embeddings. The project involves model design, implementation, and rigorous evaluation of the learned latent space's quality.

Zurück