Oral Academic Seminar about Our Work STANDS and ACSleuth
Date:
We discussed how to detect anomalous samples in multi-omics data with generative models and introduced our team’s work, ACSleuth and STANDS.
In this talk, we presented the two recent works, ACSleuth and STANDS, proposed by our team, and their applicable tasks, implementation principles, and limitations. Among them, ACSleuth is more suitable for anomalous cell detection in single-cell omics, and the specially designed MMD-based anomaly scorer enhances the algorithm’s tolerance to batch effects. On the other hand, STANDS, as a multitask framework, is better suited for spatial transcriptomics, and it can integrate gene expression with tissue imaging to accomplish various multimodal tasks.