In this project, we investigate the epigenetic landscapes of Drosophila melanogaster during embryogenesis using single-cell technologies. Specifically, we integrate data from single-cell CUT&Tag (nanoCT), single-cell RNA-seq, and single-cell ATAC-seq to study how chromatin modifications predict gene expression at the single-cell level. We are particularly interested in the mechanisms of Polycomb-mediated repression, its tissue specificity, and the formation of developmental gene-regulatory networks. Another part of the project focuses on the alternative promoter usage and epigenetic regulation of this process. We plan to use two deep learning-based tools that require GPU computing. 1. scGLUE – a variational autoencoder (VAE)-based framework for: o Integration of scATAC-seq and scCUT&Tag data, o Inference of gene-bin relationships in scRNA-seq, o Construction of gene regulatory networks [1]. 2. PRINT – a convolutional neural network (CNN)-based tool for predicting transcription factor footprints at single-cell resolution from scATAC-seq data [2]. Both tools rely on CUDA and require access to NVIDIA GPUs. As we currently lack local access to compatible GPU hardware, we are applying for access to Alvis, which would enable us to perform these analyses efficiently. We plan to analyze datasets obtained from Drosophila embryos at multiple developmental stages. In addition, we recently generated data from embryos with tissue-specific knockdown of Polycomb-mediated repression. Our analysis will also incorporate published scATAC-seq and scRNA-seq datasets. We anticipate running at least 20 separate integrations and inferences, varying inference parameters across different time points and genotypes. References: 1) Cao, ZJ., Gao, G. Nat Biotechnol (2022). https://pypi.org/project/scglue/ 2) Hu, Y., et al. Nature (2025). https://github.com/HYsxe/PRINT/tree/main