TAD Detection Based On Deep Learning

Chromosome Hi-C technology is put forward and rapid development,let people can accurately obtain the chromosome three-dimensional space position of the structure of the information,but how to find more Hi-C data with biological significance and the technical means by which to solve the problem more biological information,this is we need to focus on problems and explore the biological information technology.TAD is a region with dense interaction of gene loci on the chromosome,which contains very rich gene expression information,and the boundary also contains rich modification protein information and transcription factor information.However,the identification and detection of TAD has always been a challenging problem.Previous TAD detection methods are based on traditional statistics,and some other methods detect TAD through clustering algorithm.But their results need to be improved.In recent years,deep learning has been rapidly developed and applied in different fields,such as NLP,CV,and recommendation fields.In this article,the forefront of computer algorithm combined with biological information needs,through deep learning target detection algorithm and the fusion of traditional image processing technology,put forward a TAD domain detection method based on the deep learning,and through the test method of different organisms,different cell lines of TAD domain,the TAD domain structure information,and also by its biological significance to verify the detection algorithm of other organisms data generalization performance.The main contributions of this paper are:1)This paper explores the methods of Hi-C data preprocessing and visualization that conform to the characteristics of biological data,and proposes a close combination of HiC data and image algorithm,so that researchers can use image algorithm to explore more biological meanings of Hi-C.2)Deep TAD,a TAD domain detection algorithm based on deep learning,is proposed.The algorithm integrates the traditional image processing technology and the target detection algorithm based on deep learning,so that it can be applied to the field of biological information,and the two can be closely combined to explore the algorithm framework that can well solve the biological information problem.3)According to the characteristics of Hi-C data at the TAD domain boundary,the BC(boundary correction)algorithm is proposed to accurately modify the TAD domain boundary,so that the target detection algorithm is combined with the precise requirements of biological problems on the characteristic position,and the TAD detection result is more accurate.4)Compared with similar international methods,other biometric data were used to validate the Deep TAD results.Deep TAD test results were used to explore more biometric significance.