An Interpretable Deep Learning Approach for Biomarker Detection in LC-MS Proteomics Data

Abstract

Analyzing mass spectrometry-based proteomics data with deep learning (DL) approaches poses several challenges due to the high dimensionality, low sample size, and high level of noise. Besides, DL-based workflows are often hindered to be integrated into medical settings due to the lack of interpretable explanation. We present DLearnMS, a DL biomarker detection framework, to address these challenges on proteomics instances of liquid chromatography-mass spectrometry (LC-MS) - a well-established tool for quantifying complex protein mixtures. Our DLearnMS framework learns the clinical state of LC-MS data instances using convolutional neural networks. Next, based on the trained neural networks, biomarkers can be identified using the layer-wise relevance propagation technique. This enables detecting discriminating regions of the data and the design of more robust networks. We show that DLearnMS outperforms conventional LC-MS biomarker detection approaches in detecting fewer false positive peaks while maintaining a comparable amount of true positives peaks. Unlike other methods, no explicit preprocessing step is needed in DLearnMS.