ADEPT: Autoencoder with Differentially Expressed Genes and Imputation for a Robust Spatial Transcriptomics Clustering

Abstract: Recent advancements in spatial transcriptomics (ST) have enabled an in-depth understanding of complex tissue by allowing the measurement of gene expression at spots of tissue along with their spatial information. Several notable clustering methods have been introduced to utilize both spatial and transcriptional information in analysis of ST datasets. However, data quality across different ST sequencing techniques and types of datasets appears as a crucial factor that influences the performance of different methods and influences benchmarks. To harness both spatial context and transcriptional profile in ST data, we develop a novel graph-based multi-stage framework for robust clustering, called ADEPT. To control and stabilize data quality, ADEPT relies on selection of differentially expressed genes (DEGs) and imputation of the multiple DEG-based matrices for the initial and final clustering of a graph autoencoder backbone that minimizes the variance of clustering results. We benchmarked ADEPT against five other popular methods on ST data generated by different ST platforms. ADEPT demonstrated its robustness and superiority in different analyses such as spatial domain identification, visualization, spatial trajectory inference, and data denoising. ADEPT is freely available at GitHub: \url{https://github.com/maiziezhoulab/ADEPT}.

Download paper here

Y. Hu†, Y. Zhao†, C. T. Schunk, Y. Ma, T. Derr, X. M. Zhou. ADEPT: autoencoder with differentially expressed genes and imputation for a robust spatial transcriptomics clustering. The 13th RECOMB Satellite Conference on Biological Sequence Analysis (RECOMB-SEQ 2023).