biolord: Deciphering Single-Cell Data through Disentanglement

Researchers have introduced biolord, a deep generative model designed to disentangle single-cell multi-omic data into known and unknown attributes, such as spatial, temporal, and disease states. By separating these attributes, biolord reveals distinct biological signatures across various single-cell modalities and systems. The model can virtually transition cells across different states, enabling the generation of experimentally inaccessible samples. Notably, biolord outperforms existing methods in predicting cellular responses to previously unseen drugs and genetic perturbations, offering a powerful tool for advancing single-cell analysis.