Deep learning-based multiomics integration model for predicting axillary lymph node metastasis in breast cancer

Abstract

Aim: To develop a deep learning-based multiomics integration model. Materials & methods: Five types of omics data (mRNA, DNA methylation, miRNA, copy number variation and protein expression) were used to build a deep learning-based multiomics integration model via a deep neural network, incorporating an attention mechanism that adaptively considers the weights of multiomics features. Results: Compared with other methods, the deep learning-based multiomics integration model achieved remarkable results, with an area under the curve of 0.89 (95% CI: 0.863–0.910). Conclusion: The deep learning-based multiomics integration model achieved promising results and is an effective method for predicting axillary lymph node metastasis in breast cancer.