25 September 2020 in Macromolecules
17 March 2021 in Macromolecules
20 November 2023 in bioRxiv
Motivation Characterization of regulatory elements in DNA sequence is a key task in functional genomics. CTCF exhibits specific binding patterns in the genome of cancer cells and has a non-canonical function to facilitate oncogenic transcription program by cooperating with transcription factors bound at flanking distal regions. Identification of sequence motifs from a broad genomic region surrounding cancer-specific CTCF binding sites can help find active transcription factors in a cancer type. However, the long DNA sequences without localization information makes it difficult to perform conventional motif enrichment analysis. Results We present DNAResDualNet (DARDN), a computational method that utilizes convolutional neural networks (CNNs) coupled with feature discovery using DeepLIFT, for identifying DNA sequence features that can differentiate two sets of lengthy DNA sequences. Evaluation on DNA sequences associated with CTCF binding sites in T-cell acute lymphoblastic leukemia (T-ALL) and other cancer types demonstrates DARDN’s ability in classifying DNA sequences surrounding cancer-specific CTCF binding from control constitutive CTCF binding and identifying sequence motifs for transcription factors potentially active in each specific cancer type. We identified motifs for potential oncogenic transcription factors in T-ALL, acute myeloid leukemia (AML), breast cancer (BRCA), colorectal cancer (CRC), lung adenocarcinoma (LUAD), and prostate cancer (PRAD). Our work demonstrates the power of advanced machine learning and feature discovery approach in finding biologically meaningful information from complex DNA sequence data. Author summary We present DNAResDualNet (DARDN), a deep learning model designed for identifying DNA sequence motifs in long DNA sequences, particularly for revealing oncogenic transcription factors in various cancers. Building on our previous work (Fang et al., Genome Biol. 2020), which uncovered the non-canonical role of cancer-specific CTCF binding in oncogenesis, DARDN uses a dual convolutional neural network with residual connections and the DeepLIFT method to discern complex patterns in DNA. We successfully applied DARDN to identify unique sequence motifs in T-cell acute lymphoblastic leukemia and other cancers like AML, BRCA, CRC, LUAD, and PRAD, showcasing its capability in cancer genomics research. This work underscores the potential of deep learning in functional genomics, especially in interpreting biological data and discovering new insights in gene regulation and cancer genomics.