Biostatistics seminar
Ziang Niu, a PhD student at University of Pennsylvania, will present a talk titled "spaCRT: a computationally efficient and statistically accurate differential expression test for single-cell CRISPR screens."
About
Ziang Niu is a PhD student in the Department of Statistics and Data Science at University of Pennsylvania.
Abstract
Single-cell CRISPR screens (perturb-seq) link genetic perturbations to phenotypic changes at the resolution of individual cells. A central task in perturb-seq analysis is testing for associations between a perturbation and a count outcome, such as gene expression. These studies are typically large-scale, yet their readouts are sparse, with a high proportion of zeros. This combination poses significant challenges for existing hypothesis testing methods. Asymptotic tests are computationally efficient but often suffer from inflated Type-I error rates due to poor normal approximations. Resampling-based tests, by contrast, provide stronger Type-I error control but are computationally prohibitive at scale. To resolve this trade-off, we introduce spaCRT, a new testing procedure that leverages the classic saddlepoint approximation (SPA) to approximate the distilled conditional randomization test (dCRT). spaCRT is entirely resampling-free yet retains the finite-sample performance of dCRT, making it particularly well-suited for sparse single-cell CRISPR data. Theoretically, we establish the first rigorous result on the validity of applying SPA to resampling-based procedures. The presentation will be based on two papers:
- Z. Niu, J. Ray Choudhury & E. Katsevich (2024). Computationally efficient and statistically accurate conditional independence testing with spaCRT. Preprint.
- Z. Niu, J. Ray Choudhury & E. Katsevich (2024). The saddlepoint approximation for averages of conditionally independent random variables. Preprint.
Contact
Andy Ni