Identification of transcription factor binding sites (TFBSs) and cis-regulatory motifs (motifs for short) from genomics datasets, provides a powerful view of the rules governing the interactions between TFs and DNA. Existing motif prediction methods however, are limited by high false positive rates in TFBSs identification, contributions from non-sequence-specific binding, and complex and indirect binding mechanisms. High throughput next-generation sequencing data provides unprecedented opportunities to overcome these difficulties, as it provides multiple whole-genome scale measurements of TF binding information. Uncovering this information brings new computational and modeling challenges in high-dimensional data mining and heterogeneous data integration. To improve TFBS identification and novel motifs prediction accuracy in the human genome, we developed an advanced computational technique based on deep learning (DL) and high-performance computing, named DESSO. DESSO utilizes deep neural network and binomial distribution to optimize the motif prediction. Our results showed that DESSO outperformed existing tools in predicting distinct motifs from the 690 in vivo ENCODE ChIP-Sequencing (ChIP-Seq) datasets for 161 human TFs in 91 cell lines. We also found that protein-protein interactions (PPIs) are prevalent among human TFs, and a total of 61 potential tethering binding were identified among the 100 TFs in the K562 cell line. To further 2 expand DESSO's deep-learning capabilities, we included DNA shape features and found that (i) shape information has a strong predictive power for TF-DNA binding specificity; and (ii) it aided in identification of the shape motifs recognized by human TFs which in turn contributed to the interpretation of TF-DNA binding in the absence of sequence recognition. DESSO and the analyses it enabled will continue to improve our understanding of how gene expression is controlled by TFs and the complexities of DNA binding. The source code and the predicted motifs and TFBSs from the 690 ENCODE TF ChIP-Seq datasets are freely available at the DESSO web server: http://bmbl.sdstate.edu/DESSO. KEYWORDSTF-DNA interactions, ChIP-Seq, Motif prediction, Deep learning, DNA shape motif.DNA shape of the target sequences. Thus, features in DNA sequences in combination with shapes determine the TF binding in a more sophisticated way than was originally thought [27][28][29][30].ChIP-Seq provides the genome-wide interactions between DNA and DNA-associated proteins and large-scale ChIP-Seq efforts enable new insights into gene regulation analyses. A considerable amount of ChIP-Seq data has been generated in the public domain, including approximately 6,000 datasets of human from the ENCODE database [31]. These datasets provide an unprecedented opportunity to predict motifs, identify TFBSs, and capture more features affecting TF binding [32]. ChIP-Seq data mining and modeling have many challenges in computation, facing high-dimensional and heterogeneous data properties, and a variety of popular methods have bee...