A computational framework to dissect imputation strategies for single-cell histone modification data

Single-cell profiling of histone post-translational modifications (scHPTMs) offers a powerful lens for dissecting epigenetic regulation and cellular identity, yet low read depth and inherent noise in these datasets pose significant analytical challenges. Here, we introduce the first comprehensive computational framework that systematically evaluates imputation strategies on scHPTM data, including methods originally developed for scRNA-seq and scATAC-seq. Leveraging both synthetic and published datasets, we apply novel performance metrics-implemented in a modular R package-to assess signal recovery, enrichment at biologically relevant genomic sites, and preservation of cell-to-cell similarities. Our extensive benchmarking reveals that performance varies markedly by analytical task (e.g. signal denoising, peak detection, and clustering), highlighting that no one-size-fits-all solution exists for these data. By delineating the strengths and limitations of current imputation approaches, this work lays the foundation for the targeted development of next-generation, task-aware algorithms, while providing critical guidance for researchers and developers on the current capabilities and unmet needs in single-cell epigenomics.