Deep Neural Networks for Classification of LC-MS Spectral Peaks

Abstract

Liquid chromatography−mass spectrometry (LC-MS)based metabolomics has emerged as a valuable tool for biological discovery, capable of assaying thousands of diverse chemical entities in a single biospecimen. Processing of nontargeted LC-MS spectral data requires identification and isolation of true spectral features from the random, false noise peaks that comprise a significant portion of total signals, using inexact peak selection algorithms and time-consuming visual inspection of data. To increase the fidelity and speed of data processing, herein we establish, optimize, and evaluate a machine learning pipeline employing deep neural networks as well as a simpler multiple logistic regression model for classification of spectral features from nontargeted LC-MS metabolomics data. Machine learning-based approaches were found to remove up to 90% of false peaks from complex nontargeted LC-MS data sets without reducing true positive signals and exhibit excellent reproducibility across multiple data sets. Application of machine learning for nontargeted LC-MS-based peak selection provides for robust and scalable peak classification and data filtering, enabling handling and processing of large scale, complex metabolomics data sets.

Decreased adipose tissue oxygenation associates with insulin resistance in individuals with obesity

“…The homeostasis model assessment of insulin resistance (HOMA-IR) was calculated as the product of fasting plasma insulin (in μU/mL) and glucose (in mmol/L) concentrations divided by 22.5 (76). Plasma BCAA concentrations were determined using liquid chromatography-tandem mass spectrometry as previously described (77). Plasma cytokine and chemokine concentrations were determined using commercially available magnetic bead suspension assays (MilliporeSigma and R&D Systems) and a Luminex 200 analyzer (Luminex Corp.).…”

Deep metabolome: Applications of deep learning in metabolomics

“…DL based peak filtering approaches seek to overcome the limitations of traditional methods for handling low signal to noise, diverse and irregular peak shapes and poor baseline resolution. For example, Kantz et al [56] used a CNN model to detect true spectral peaks vs. artifacts using stacked peak images of LC-MS chromatographic features as input data. This approach was shown to eliminate up to 90% of all false noise peaks.…”

Deep learning in systems medicine

“…Variational autoencoders [82] and Generative Adversarial Networks [83] approaches adopted in SM are increasingly multiscale [95] and the data processing workflows consist of a multi-step strategy involving various chemometrics and bioinformatics tools [96] in which DL has recently brought new horizons. As an example, DNN has been used for spectral peak classification in the development of several tools that improve data extraction [97,98]. A DNN-MDA approach has also been shown of interest in determining important variables in complex datasets, in the context of biomarker discovery [99].…”