A machine learning model for CUT&Tag bias correction

Accurate detection of transcription factor (TF) binding sites and histone modifications (HM) on the genome-wide scale is essential for studying epigenome functions and transcriptional regulation. Cleavage Under Targets & Tagmentation (CUT&Tag) is a low-cost, easy-to-implement epigenomic profiling method that can be performed on a small number of cells and at the single-cell level. CUT&Tag experiments use the hyperactive transposase Tn5 for tagmentation. However, we find that Tn5 exhibits intrinsic sequence insertion bias. Furthermore, preference of Tn5 insertion toward accessible chromatin influences the distribution of CUT&Tag reads. Such intrinsic sequence bias and open chromatin bias can significantly confound both bulk and single-cell CUT&Tag data analysis, which requires careful assessment and new analytical methods. To address this challenge, we present PATTY (Propensity Analyzer for Tn5 Transposase Yielded biases), a computational method for systematic characterization and correction of biases in CUT&Tag data. Our results show that histone modification signals detected from CUT&Tag, after bias correction, are better associated with orthogonal biological features such as gene expression. Additionally, single-cell clustering using the corrected single-cell CUT&Tag signals better reflects cell type specificity. This new computational method can enhance the analysis of both bulk and single-cell CUT&Tag data.