ProteoNet: A CNN-based framework for analyzing proteomics MS-RGB images

• DOI: 10.1016/j.isci.2024.111362
• Publication Date: 2024-11-12T21:05:52Z
• Authors: Jinze Huang; Yimin Li; Bo Meng; Yong Zhang; Yaoguang Wei; Xinhua Dai; Dong An; Yang Zhao; Xiang Fang
• Keywords: RGB color model

Proteomics is crucial in clinical research, yet the application of proteomic data remains challenging. Transforming mass spectrometry (MS) into red, green, and blue color (MS-RGB) image formats applying deep learning (DL) techniques has shown great potential to enhance analysis efficiency. However, current DL models often fail extract subtle, features from MS-RGB data. To address this, we developed ProteoNet, a framework that refines analysis. ProteoNet incorporates semantic partitioning, adaptive average pooling, weighted factors Convolutional Neural Network (CNN) model, thus enhancing accuracy. Our experiments with proteomics urine, blood, tissue samples related liver, kidney, thyroid diseases demonstrate outperforms existing also provides direct conversion method for data, enabling seamless workflow. Moreover, its compatibility various CNN architectures, including lightweight like MobileNetV2, underscores scalability potential.