gao-lab/GLUE

Implements a variational graph autoencoder framework that jointly embeds multiple omics modalities (RNA, ATAC, protein) by leveraging prior biological networks as graph constraints during training. Integrates with the AnnData ecosystem and PyTorch, enabling GPU-accelerated inference on large-scale single-cell datasets with built-in evaluation pipelines for cross-modality matching and clustering assessment.

456 stars.