Deep Learning on Abnormal Chromosome Segments: An Intelligent Copy Number Variants Detection System Design

Gene testing emerged as a business in the last two decades, and the testing cost has been reduced from 100 million to 1000 dollars for the development of technologies. Preimplantation genetic screening (PGS) is a popular genetic profiling of embryos prior to implantation in gene testing. Copy number variants (CNVs) detection is a key task in PGS which still needs the manual operation and evaluation. At the same time, deep learning technology earns a booming development and wide application in recent years for its strong computing and learning capability. This research redesigns the PGS workflow with the intelligent CNVs detection system, and proposes the corresponding system framework. Deep learning is selected as the proper technology in the system design for CNVs detection, which also fit the task of denoising. The evaluation is conducted on simulation dataset with high accuracy and low time cost, which may achieve the requirements of clinical application and reduce the workload of bioinformatics experts. Moreover, the redesigned process and proposed framework may enlighten the intelligent system design for gene testing in following work, and provide a guidance of deep learning application in AI healthcare.