No abstract

No abstract

Motivation: The Brain and Muscle ARNTL-Like 1 protein (BMAL1) forms a heterodimer with either Circadian Locomotor Output Cycles Kaput (CLOCK) or Neuronal PAS domain protein 2 (NPAS2) to act as a master regulator of the mammalian circadian clock gene network. The dimer binds to E-box gene regulatory elements, activating downstream transcription of clock genes. Identification of transcription factor binding sites and features that correlate to DNA binding by BMAL1 is a challenging problem, given that CLOCK-BMAL1 or NPAS2-BMAL1 bind to several distinct binding motifs (CANNTG) on DNA. Results: Using three different types of tissue-specific machine learning models with features based on 1) DNA sequence, 2) DNA sequence plus DNA shape, and 3) DNA sequence and shape plus histone modifications, we developed an interpretable predictive model of genome-wide BMAL1 binding to E-box motifs and dissected the mechanisms underlying BMAL1-DNA binding. Our results indicated that histone modifications, the local shape of the DNA, and the flanking sequence of the E-box motif are sufficient predictive features for BMAL1-DNA binding. Our models also provide mechanistic insights into tissue specificity of DNA binding by BMAL1.

The mammalian circadian clock is based on a core intracellular gene regulatory network, coordinated by communication between the central nervous system and peripheral tissues like the liver. Transcriptional and translational feedback loops underlie the molecular mechanism of circadian oscillation and generate its 24 h periodicity. The Brain and muscle Arnt-like protein-1 (Bmal1) forms a heterodimer with Circadian Locomotor Output Cycles Kaput (Clock) that binds to E-box gene regulatory elements, activating transcription of clock genes. In this work we aimed to develop a predictive model of genome-wide CLOCK-BMAL1 binding to E-box motifs. We found over-representation of the canonical E-box motif CACGTG in BMAL1-bound regions in accessible chromatin of the mouse liver, heart and kidney. We developed three different tissue-specific machine learning models based on DNA sequence, DNA sequence plus DNA shape, and DNA sequence and shape plus histone modifications. Combining DNA sequence with DNA shape and histone modification features yielded improved transcription factor binding site prediction. Further, we identified the genomic and epigenomic features that best correlate to the binding of BMAL1 to DNA. The DNA shape features Electrostatic Potential, Minor Groove Width and Propeller Twist together with the histone modifications H3K27ac, H3K4me1, H3K36me3, and H3K4me3 were the features most highly predictive of DNA binding by BMAL1 across all three tissues.

No abstract