基于流动注意力机制的非侵入性癌症检测

Non-invasive cancer detection based on flow attention mechanism

在癌症检测领域,细胞游离DNA的高通量测序技术已引发一场重大变革,为非侵入性癌症检测提供了新的可能性。利用测序数据做出可靠且精确的预测至关重要,但是测序成本高昂。针对这一需求,提出一种基于流动注意力机制的深度学习模型。通过定义差异甲基化区域对数据进行预处理,使得满足深度学习数据量的要求,并整合全基因组双硫酸盐测序数据中的DNA序列和甲基化信息,以实现对髓母细胞瘤患者进行预测。实验结果表明,该方法提高了诊断过程的准确性,且受试者工作特征曲线面积达到99.73%,展示了深度学习技术在癌症早期诊断中的潜在应用前景。

In the field of cancer detection,the high-throughput sequencing technology of cell-free DNA has triggered a significant revolution,providing new possibilities for non-invasive cancer detection methods.Using sequencing data to make reliable and accurate predictions is critical,but sequencing is expensive.To address this demand,a deep learning model based on dynamic attention mechanism is proposed.Data were preprocessed by defining differential methylation regions to meet deep learning data volume requirements,and DNA sequence and methylation information from whole-genome disulfate sequencing data were integrated to achieve prediction of medulloblastoma patients.The experimental results show that the method can improve the accuracy of the diagnosis process,and the area under the subject working characteristic curve can reach 99.73%,demonstrating the potential application prospect of deep learning technology in the early diagnosis of cancer.