DeepIso: A Deep Learning Model for Peptide Feature Detection from LC-MS map

Abstract

Liquid chromatography with tandem mass spectrometry (LC-MS/MS) based quantitative proteomics provides the relative different protein abundance in healthy and disease-afflicted patients, which offers the information for molecular interactions, signaling pathways, and biomarker identification to serve the drug discovery and clinical research. Typical analysis workflow begins with the peptide feature detection and intensity calculation from LC-MS map. We are the first to propose a deep learning based model, DeepIso, that combines recent advances in Convolutional Neural Network (CNN) and Recurrent Neural Network (RNN) to detect peptide features of different charge states, as well as, estimate their intensity. Existing tools are designed with limited engineered features and domain-specific parameters, which are hardly updated despite a huge amount of new coming proteomic data. On the other hand, DeepIso consisting of two separate deep learning based modules, learns multiple levels of representation of high dimensional data itself through many layers of neurons, and adaptable to newly acquired data. The peptide feature list reported by our model matches with 97.43% of high quality MS/MS identifications in a benchmark dataset, which is higher than the matching produced by several widely used tools. Our results demonstrate that novel deep learning tools are desirable to advance the state-of-the-art in protein identification and quantification.

Deciphering Human Leukocyte Antigen Susceptibility Maps From Immunopeptidomics Characterization in Oncology and Infections

“…DeepIso consists of two different deep learning-based components, which can learn multiple levels of the high-dimensional data itself represented by multiple layers of neurons and can be adapted to new situations. The peptide characteristics list investigated with this model matches with 97.43% of high quality MS/MS identifications in a standard dataset (Zohora et al, 2019). The third one is DeepNovo-DIA, de novo peptide sequencing Independent data acquisition (DIA) method of mass spectrometry data.…”

Machine Learning in Mass Spectrometric Analysis of DIA Data

“…These encouraging reports lead us to propose an exciting vision of a potential path forward for DIA data analysis, which is to discard even more levels of manual intervention, and to create a method that can allow for direct operation on the raw LC/MS‐MS maps. One research group has shown some applications for this way forward for the purposes of isotopic pattern detection in DDA data analysis and DIA de novo sequencing; DeepISO [ 21 ] and DeepNovo‐DIA [ 22 ] both operate directly on data from the LC/MS‐MS feature maps for their respective tasks. Although these methods still require a degree of pre‐ and post‐processing, they show that it is possible to automate increasingly large portions of the workflow that previously required large amounts of human expertise and prior knowledge.…”

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

“…The success of deep learning (DL)-based methods, often replacing state-of-the-art classical model-based methods, in many fields such as medical imaging [17], biomedicine [18], and healthcare [19], has also encouraged the use of DL models for LC-MS proteomics analysis. To name a few, DeepIso [20] that combines a convolutional neural network (CNN) with a recurrent neural network (RNN) to detect peptide features; DeepNovo [21] and DeepNovo-DIA [22], which use DL-based approach (CNN coupled with RNN) for peptide sequencing on data-dependent acquisition (tandem-mass spectra) and data-independent acquisition MS data, respectively; pDeep [23] that adapt the bidirectional long short term memory for the spectrum prediction of peptides; and DeepRT [24] that employs a capsule network to predict RT by learning features of embedded amino acids in peptides. Despite the current successful DL approaches on analyzing LC-MS proteomics, most of the studies are empirically driven, and having a justifiable interpretation foundation is largely missing [25].…”