Oral Academic Presentation about Our Work ACSleuth

We formally introduced our work, Domain Adaptive and Fine-grained Anomaly Detection for Single-cell Sequencing Data and Beyond.

In this talk, We discovered and mathematically theorized that domain shifts prevalent in multi-sample and multi-domain datasets can impact the accuracy of anomaly detection. We also provided new insights into the limitations of existing anomaly detection methods. Moreover, inspired by the theories about Maximum Mean Discrepancy and Ramanujan’s master theorem, we proposed an innovative framework, ACSleuth, for discovering anomalous cell types and subtypes in single-cell sequencing data. Extensive experiments and theoretical analysis demonstrate its superiority over existing methods and robustness to domain shifts. More information can be found at ACSleuth.