Search

Abstract Identifying cell-type-specific 3D chromatin interactions between regulatory elements can help decipher gene regulation and interpret disease-associated non-coding variants. However, achieving this resolution with current 3D genomics technologies is often infeasible given limited input cell numbers. We therefore present ChromaFold, a deep learning model that predicts 3D contact maps, including regulatory interactions, from single-cell ATAC sequencing (scATAC-seq) data alone. ChromaFold uses pseudobulk chromatin accessibility, co-accessibility across metacells, and a CTCF motif track as inputs and employs a lightweight architecture to train on standard GPUs. Trained on paired scATAC-seq and Hi-C data in human samples, ChromaFold accurately predicts the 3D contact map and peak-level interactions across diverse human and mouse test cell types. Compared to leading contact map prediction models that use ATAC-seq and CTCF ChIP-seq, ChromaFold achieves state-of-the-art performance using only scATAC-seq. Finally, fine-tuning ChromaFold on paired scATAC-seq and Hi-C in a complex tissue enables deconvolution of chromatin interactions across cell subpopulations.

Abstract Acute myeloid leukemia (AML) is a high-risk malignancy characterized by a diverse spectrum of somatic genetic alterations. The mechanisms by which these mutations contribute to leukemia development and how this informs the use of targeted therapies is critical to improving outcomes for patients. Importantly, how to target loss-of-function mutations has been a critical challenge in precision medicine. Heterozygous inactivating mutations in cohesin complex genes contribute to AML in adults by increasing the self-renewal capacity of hematopoietic stem and progenitor cells (HSPCs) by altering PRC2 targeting to induce HOXA9 expression, a key self-renewal transcription factor. Here we sought to delineate the epigenetic mechanism underpinning the enhanced self-renewal conferred by cohesin-haploinsufficiency. First, given the substantial difference in the mutational spectrum between pediatric and adult AML patients, we first sought to identify if HOXA9 was also elevated in children. Next, using primary HSPCs as a model we demonstrate that abnormal self-renewal due to cohesin loss is blocked by DOT1L inhibition. In cohesin-depleted cells, DOT1L inhibition is associated with H3K79me2 depletion and a concomitant increase in H3K27me3. Importantly, we find that there are cohesin-dependent gene expression changes that promote a leukemic profile, including HoxA overexpression, that are preferentially reversed by DOT1L inhibition. Our data further characterize how cohesin mutations contribute to AML development, identifying DOT1L as a potential therapeutic target for adult and pediatric AML patients harboring cohesin mutations.