RESUMEN
J-Resolved (J-Res) nuclear magnetic resonance (NMR) spectroscopy is pivotal in NMR-based metabolomics, but practitioners face a choice between time-consuming high-resolution (HR) experiments or shorter low-resolution (LR) experiments which exhibit significant peak overlap. Deep learning neural networks have been successfully used in many fields to enhance quality of natural images, especially with regard to resolution, and therefore offer the prospect of improving two-dimensional (2D) NMR data. Here, we introduce the J-RESRGAN, an adapted and modified generative adversarial network (GAN) for image super-resolution (SR), which we trained specifically for metabolomic J-Res spectra to enhance peak resolution. A novel symmetric loss function was introduced, exploiting the inherent vertical symmetry of J-Res NMR spectra. Model training used simulated high-resolution J-Res spectra of complex mixtures, with corresponding low-resolution spectra generated via blurring and down-sampling. Evaluation of peak pair resolvability on J-RESRGAN demonstrated remarkable improvement in resolution across a variety of samples. In simulated plasma data, 100% of peak pairs exhibited enhanced resolution in super-resolution spectra compared to their low-resolution counterparts. Similarly, enhanced resolution was observed in 80.8-100% of peak pairs in experimental plasma, 85.0-96.7% in urine, 94.4-98.9% in full fat milk, and 82.6-91.7% in orange juice. J-RESRGAN is not sample type, spectrometer or field strength dependent and improvements on previously acquired data can be seen in seconds on a standard desktop computer. We believe this demonstrates the promise of deep learning methods to enhance NMR metabolomic data, and in particular, the power of J-RESRGAN to elucidate overlapping peaks, advancing precision in a wide variety of NMR-based metabolomics studies. The model, J-RESRGAN, is openly accessible for download on GitHub at https://github.com/yanyan5420/J-RESRGAN.
Asunto(s)
Aprendizaje Profundo , Espectroscopía de Resonancia Magnética , Metabolómica , Metabolómica/métodos , Espectroscopía de Resonancia Magnética/métodos , Animales , HumanosRESUMEN
MOTIVATION: Random forests (RFs) can deal with a large number of variables, achieve reasonable prediction scores, and yield highly interpretable feature importance values. As such, RFs are appropriate models for feature selection and further dimension reduction. However, RFs are often not appropriate for correlated datasets due to their mode of selecting individual features for splitting. Addressing correlation relationships in high-dimensional datasets is imperative for reducing the number of variables that are assigned high importance, hence making the dimension reduction most efficient. Here, we propose the LAtent VAriable Stochastic Ensemble of Trees (LAVASET) method that derives latent variables based on the distance characteristics of each feature and aims to incorporate the correlation factor in the splitting step. RESULTS: Without compromising on performance in the majority of examples, LAVASET outperforms RF by accurately determining feature importance across all correlated variables and ensuring proper distribution of importance values. LAVASET yields mostly non-inferior prediction accuracies to traditional RFs when tested in simulated and real 1D datasets, as well as more complex and high-dimensional 3D datatypes. Unlike traditional RFs, LAVASET is unaffected by single 'important' noisy features (false positives), as it considers the local neighbourhood. LAVASET, therefore, highlights neighbourhoods of features, reflecting real signals that collectively impact the model's predictive ability. AVAILABILITY AND IMPLEMENTATION: LAVASET is freely available as a standalone package from https://github.com/melkasapi/LAVASET.