Machine Learning Researches On 3D Structure Of Chromatin And Prognostic Model Of DNA Methylation

According to the research results of the World Health Organization(WHO),cancer is still a fatal disease with high morbidity and mortality.The causes of malignant tumors are various,including abnormal expression of proto-oncogenes,chromosomal structural variation(SV)and the silencing of tumor suppressor gene transcription.Early screening and diagnosis of cancer is an important problem that plagues scientific researchers and medical scientists.Traditional methods such as high-throughput chromosome conformation capture(Hi-C)and optical mapping can only provide limited sequence information and resolution.However,based on existing biological data to expand the visualization of chromatin tissue conformation in high-dimensional space is still lack of research strategies.Abnormal DNA methylation levels are often early warning signs of cancer.Therefore,how to identify high risks Methylation sites,combined with clinical data,gene expression data,etc.,to provide more rigorous and rich prognostic information,still need new research programs.In this article,aiming at the research of chromatin spatial conformation and prognostic analysis of DNA methylation,the biological theories and machine learning algorithms are combined to explore the potential characteristics of biological data,and explore the research methods that best match the law of data samples,and for cancer Issues and other bioinformatics issues to provide ideas,the main contributions of this article are as follows:(1)Aiming at the problem of chromatin spatial imaging,originally proposed a research framework of three-dimensional chromatin conformation based on Hilbert Curve.Using the dinucleotide-based gene sequence feature extraction method,extract the gene sequence feature,combine the Hi-C data,use the local weighted linear regression(LWR)machine learning algorithm to fit the Hilbert key value,and then restore the Hilbert key value to three-dimensional space.Realize the visualization of chromatin conformation and provide a new method for the study of chromatin spatial structure.(2)Propose a DNA methylation prognostic analysis model based on the PADMXB algorithm for DNA methylation data set and least squares linear regression.Aiming at the problem of unbalanced DNA methylation data set samples,threshold parameters are added to the model,and the classification boundary of positive and negative samples is adjusted.Compared with Cox regression analysis,the algorithm effect is improved.Combining clinical data,using least squares linear regression algorithm to fit survival time,through the combined analysis of selected methylation sites and gene expression,in-depth research on cancer problems.