GFETM: Genome Foundation-based Embedded Topic Model for scATAC-seq Modeling

Single-cell Assay for Transposase-Accessible Chromatin with sequencing (scATAC-seq) has emerged as a powerful technique for investigating open chromatin landscapes at the single-cell level. Yet, scATAC-seq cell representation learning and its downstream tasks remain challenging due to the inherent high dimensional, sparse, and noisy properties of the data. The scarcity of available datasets compared to scRNA-seq further underscores the importance of applying transfer learning from abundant reference data to enhance scATAC-seq analyses across diverse biological scenarios. However, variations in computational methods and inherent biological differences between scATAC-seq samples intensify the difficulty in effectively implementing transfer learning strategies. Genome Foundation Models (GFMs), which are pre-trained on millions of DNA sequences in an self-supervised manner via masked nucleotide prediction, have proven effective in applications involving genomic sequences, yet their application in single-cell biology remains underexplored. Given that highly accessible chromatin regions often harbour salient sequence features, we hypothesize that leveraging GFM nucleotide sequence embeddings may improve scATAC-seq data modeling and its transferability. In this study, we introduce the Genome Foundation Embedded Topic Model (GFETM), an interpretable and transferable deep neural network framework that combines GFMs with the Embedded Topic Model (ETM) for scATAC-seq data analysis. We show that by probing and integrating the DNA sequence embedding extracted by GFMs from open chromatin regions, GFETM not only achieves state-of-the-art performance of scATAC-seq cell representation learning on benchmarking datasets of various scales but also demonstrates generalizability and transferability to single-cell transcriptomes and across different subjects, tissues, and species. ### Competing Interest Statement The authors have declared no competing interest.