Machine Learning in Mass Spectrometric Analysis of DIA Data

Abstract

Liquid Chromatography coupled to Tandem Mass Spectrometry (LC-MS/MS) based methods are currently the top choice for high-throughput, quantitative measurements of the proteome. While traditional proteomics LC-MS/MS methods can suffer from issues such as low reproducibility and quantitative accuracy due to its stochastic nature, recent improvements in acquisition protocols have resulted in methods that can overcome these challenges. Data-independent acquisition (DIA) is a novel mass spectrometric method that does so by using a deterministic acquisition strategy. These new approaches will allow researchers to apply MS on more complex samples, however, existing heuristic and expert-knowledge based methods will struggle with keeping pace of the increasing complexity of the resulting data. Deep learning (DL) based methods have been shown to be more adept at handling large amounts of complex data than traditional methods in many other fields, such as computer vision and natural language processing. Proteomics is also entering a phase where the size and complexity of the data will require us to look towards scalable and data-driven DL pipelines.

Computational Proteomics: Focus on Deep Learning

“…[10] In the third article, Xu et al summarize deep learning applications in the DIA data analysis workflow and discuss the interpretability of deep learning models. [11] The third group of articles describes methods and tools for multi-omics data analysis. In the first article, Calinawan et al present a web application called ProTrack, which allows researchers to intuitively query, explore, and download data and analysis results from the CPTAC projects.…”

Deep Learning in Proteomics

“…Due to these computational requirements, machine learning methods have been widely used in many aspects of proteomics data analysis. [1][2][3] Deep learning is a sub-discipline of machine learning. It has advanced rapidly during the last two decades and has demonstrated superior performance in various fields including computer vision, speech recognition, natural-language processing, bioinformatics, and medical image analysis.…”

AHLF: ad hoc learning of peptide fragmentation from mass spectra enables an interpretable detection of phosphorylated and cross-linked peptides

“…In reality, this is only met when the architecture has a build-in trainable redundancy that matches the nature of the problem. Also, the learning process including the used tensor operations have to be efficient enough to cope with the vast amount of training data LeCun et al (2015); Xu et al (2020). Yet, for complex and high-dimensional data points this is mostly ever successful when parameter sharing techniques such as convolutional or recurrent neural networks are used.…”

“…A generic end-to-end training of deep learning models directly applied to tandem mass spectra could greatly benefit the proteomics community Xu et al (2020). Here, we developed (i) a loss-less spectrum representation that allows efficient model training and (ii) set up a capable model architecture that accounts for the high resolution and complex nature of tandem mass spectra.…”