maxATAC

maxATAC predicts transcription factor (TF) binding sites genome-wide from Assay-for-Transposase-Accessible-Chromatin sequencing (ATAC-seq) data to support mapping gene regulatory networks and chromatin accessibility–associated regulation.

Key Features:

• Deep Neural Network Models: Uses deep neural network architectures specifically tailored for TFBS prediction from ATAC-seq and reports improved performance relative to traditional motif scanning methods.
• Extensive Benchmark Dataset: Trains and validates models on a curated benchmark dataset comprising 127 human transcription factors.
• Support for Single-cell and Primary Cells: Applies to bulk ATAC-seq, primary cell data, and single-cell ATAC-seq for TFBS prediction across diverse cellular contexts.
• Comprehensive Model Collection: Provides a large collection of high-performance models covering 127 human transcription factors for ATAC-seq–based binding prediction.

Scientific Applications:

• Gene regulatory network reconstruction: Enables identification of TF binding sites to inform reconstruction of gene regulatory networks across cell types.
• Single-cell regulatory analysis: Facilitates TFBS prediction at single-cell resolution from single-cell ATAC-seq data to study cell-to-cell regulatory variation.
• Disease locus interpretation: Has been used to identify TFBS associated with allele-dependent chromatin accessibility at genetic risk loci for atopic dermatitis.
• Developmental and precision medicine studies: Supports analysis of TF binding dynamics relevant to developmental biology and precision medicine investigations.

Methodology:

Integrates deep neural network architectures trained and validated on a curated benchmark dataset of 127 human TFs to predict TFBS from ATAC-seq data, with performance contrasted against traditional motif scanning approaches.