Mass Spectrometry Imaging Datasets Research Articles

Mass spectrometry imaging-based multi-modal technique: Next-generation of biochemical analysis strategy

Mass spectrometry imaging-based multi-modal technique: Next-generation of biochemical analysis strategy

Contrast optimization of mass spectrometry imaging (MSI) data visualization by threshold intensity quantization (TrIQ).

Mass spectrometry imaging (MSI) enables the unbiased characterization of surfaces with respect to their chemical composition. In biological MSI, zones with differential mass profiles hint towards localized physiological processes, such as the tissue-specific accumulation of secondary metabolites, or diseases, such as cancer. Thus, the efficient discovery of ‘regions of interest’ (ROI) is of utmost importance in MSI. However, often the discovery of ROIs is hampered by high background noise and artifact signals. Especially in ambient ionization MSI, unmasking biologically relevant information from crude data sets is challenging. Therefore, we implemented a Threshold Intensity Quantization (TrIQ) algorithm for augmenting the contrast in MSI data visualizations. The simple algorithm reduces the impact of extreme values (‘outliers’) and rescales the dynamic range of mass signals. We provide an R script for post-processing MSI data in the imzML community format (https://bitbucket.org/lababi/msi.r) and implemented the TrIQ in our open-source imaging software RmsiGUI (https://bitbucket.org/lababi/rmsigui/). Applying these programs to different biological MSI data sets demonstrated the universal applicability of TrIQ for improving the contrast in the MSI data visualization. We show that TrIQ improves a subsequent detection of ROIs by sectioning. In addition, the adjustment of the dynamic signal intensity range makes MSI data sets comparable.

RMSIannotation: A peak annotation tool for mass spectrometry imaging based on the analysis of isotopic intensity ratios

Mass spectrometry imaging (MSI) consist of spatially located spectra with thousands of peaks. Only a fraction of these peaks corresponds to unique monoisotopic peaks, as mass spectra include isotopes, adducts and fragments of compounds. Current peak annotation solutions depend on matching MS features to compounds libraries. We present rMSIannotation, a peak annotation algorithm to annotate carbon isotopes and adducts in metabolomics and lipidomics imaging mass spectrometry datasets without using supporting libraries. rMSIannotation measures and evaluates the intensity ratio between carbon isotopic peaks and models their distribution across the m/z axis of the compounds in the Human Metabolome Database. Monoisotopic peak selection is based on the isotopic likelihood score (ILS) made of three components: image morphology correlation, validation of isotopic intensity ratios, and peak centroid mass deviation. rMSIannotation proposes pairs of peaks that can be adducts based on three scores: isotopic pattern coherence, image correlation and mass error. We validated rMSIannotation with three MALDI-MSI datasets which were manually annotated by experts, and compared the annotations obtained with rMSIannotation and with the METASPACE annotation platform. rMSIannotation replicated more than 90% of the manual annotation reported in FT-ICR datasets and expanded the list of annotated compounds with additional monoisotopic peaks and neutral masses. Finally, we evaluated isotopic peak annotation as a data reduction method for MSI by comparing the results of PCA and k-means segmentation before and after removing non-monoisotopic peaks. The results show that monoisotopic peaks retain most of the biologic variance in the dataset.

Data Filtering and Its Prioritization in Pipelines for Spatial Segmentation of Mass Spectrometry Imaging

Mass spectrometry imaging (MSI) could provide vast amounts of data at the temporal-spatial scale in heterogeneous biological specimens, which challenges us to segment accurately suborgans/microregions from complex MSI data. Several pipelines had been proposed for MSI spatial segmentation in the past decade. More importantly, data filtering was found to be an efficient procedure to improve the outcomes of MSI segmentation pipelines. It is not clear, however, how the filtering procedure affects the MSI segmentation. An improved pipeline was established by elaborating the filtering prioritization and filtering algorithm. Lipidomic-characteristic-based MSI data of a whole-body mouse fetus was used to evaluate the established pipeline on localization of the physiological position of suborgans by comparing with three commonly used pipelines and commercial SCiLS Lab software. Two structural measurements were used to quantify the performances of the pipelines including the percentage of abnormal edge pixel (PAEP) and CHAOS. Our results demonstrated that the established pipeline outperformed the other pipelines in visual inspection, spatial consistence, time-cost, and robustness analysis. For example, the dorsal pallium (isocortex) and hippocampal formation (Hpf) regions, midbrain, cerebellum, and brainstem on the mouse brain were annotated and located by the established pipeline. As a generic pipeline, the established pipeline could help with the accurate assessment and screening of drug/chemical-induced targeted organs and exploration of the progression and molecular mechanisms of diseases. The filter-based strategy is expected to become a critical component in the standard operating procedure of MSI data sets.

Spatially aware clustering of ion images in mass spectrometry imaging data using deep learning

Computational analysis is crucial to capitalize on the wealth of spatio-molecular information generated by mass spectrometry imaging (MSI) experiments. Currently, the spatial information available in MSI data is often under-utilized, due to the challenges of in-depth spatial pattern extraction. The advent of deep learning has greatly facilitated such complex spatial analysis. In this work, we use a pre-trained neural network to extract high-level features from ion images in MSI data, and test whether this improves downstream data analysis. The resulting neural network interpretation of ion images, coined neural ion images, is used to cluster ion images based on spatial expressions. We evaluate the impact of neural ion images on two ion image clustering pipelines, namely DBSCAN clustering, combined with UMAP-based dimensionality reduction, and k-means clustering. In both pipelines, we compare regular and neural ion images from two different MSI datasets. All tested pipelines could extract underlying spatial patterns, but the neural network-based pipelines provided better assignment of ion images, with more fine-grained clusters, and greater consistency in the spatial structures assigned to individual clusters. Additionally, we introduce the relative isotope ratio metric to quantitatively evaluate clustering quality. The resulting scores show that isotopical m/z values are more often clustered together in the neural network-based pipeline, indicating improved clustering outcomes. The usefulness of neural ion images extends beyond clustering towards a generic framework to incorporate spatial information into any MSI-focused machine learning pipeline, both supervised and unsupervised.Graphical abstract

Adaptive Pixel Mass Recalibration for Mass Spectrometry Imaging Based on Locally Endogenous Biological Signals.

Mass spectrometry imaging (MSI) is a powerful and convenient method for revealing the spatial chemical composition of different biological samples. Molecular annotation of the detected signals is only possible if a high mass accuracy is maintained over the entire image and the m/z range. However, the change in the number of ions from pixel-to-pixel of the biological samples could lead to small fluctuations in the detected m/z-values, called mass shift. The use of internal calibration is known to offer the best solution to avoid, or at least to reduce, mass shifts. Their "a priori" selection for a global MSI acquisition is prone to false positive detection and therefore to poor recalibration. To fill this gap, this work describes an algorithm that recalibrates each spectrum individually by estimating its mass shift with the help of a list of pixel-specific internal calibrating ions, automatically generated in a data-adaptive manner (https://github.com/LaRoccaRaphael/MSI_recalibration). Through a practical example, we applied the methodology to a zebrafish whole-body section acquired at a high mass resolution to demonstrate the impact of mass shift on data analysis and the capability of our algorithm to recalibrate MSI data. In addition, we illustrate the broad applicability of the method by recalibrating 31 different public MSI data sets from METASPACE from various samples and types of MSI and show that our recalibration significantly increases the numbers of METASPACE annotations (gaining from 20 up to 400 additional annotations), particularly the high-confidence annotations with a low false discovery rate.

Spatial Segmentation of Mass Spectrometry Imaging Data by Combining Multivariate Clustering and Univariate Thresholding.

Spatial segmentation partitions mass spectrometry imaging (MSI) data into distinct regions, providing a concise visualization of the vast amount of data and identifying regions of interest (ROIs) for downstream statistical analysis. Unsupervised approaches are particularly attractive, as they may be used to discover the underlying subpopulations present in the high-dimensional MSI data without prior knowledge of the properties of the sample. Herein, we introduce an unsupervised spatial segmentation approach, which combines multivariate clustering and univariate thresholding to generate comprehensive spatial segmentation maps of the MSI data. This approach combines matrix factorization and manifold learning to enable high-quality image segmentation without an extensive hyperparameter search. In parallel, some ion images inadequately represented in the multivariate analysis were treated using univariate thresholding to generate complementary spatial segments. The final spatial segmentation map was assembled from segment candidates that were generated using both techniques. We demonstrate the performance and robustness of this approach for two MSI data sets of mouse uterine and kidney tissue sections that were acquired with different spatial resolutions. The resulting segmentation maps are easy to interpret and project onto the known anatomical regions of the tissue.

Structure-Preserving and Perceptually Consistent Approach for Visualization of Mass Spectrometry Imaging Datasets.

Mass spectrometry imaging (MSI) has become an important tool for 2D profiling of biological tissues, allowing for the visualization of individual compound distributions in the sample. Based on this information, it is possible to investigate the molecular organization within any particular tissue and detect abnormal regions (such as tumor regions) and many other biologically relevant phenomena. However, the large number of compounds present in the spectra hinders the productive analysis of large MSI datasets when utilizing standard tools. The heterogeneity of samples makes exploratory visualization (a presentation of the general idea of the molecular and structural organization of the inspected tissues) challenging. Here, we explore the application of various dimensionality reduction techniques that have been used extensively in the visualization of hyperspectral images and the MSI data specifically, such as principal component analysis, independent component analysis, non-negative matrix factorization, t-distributed stochastic neighbor embedding, and uniform manifold approximation and projection. Further, we propose a new approach based on a combination of structure preserving visualization with nonlinear manifold embedding of normalized spectral data. This way, we aim to preserve as much spatially overlapping signals as possible while augmenting them with information on compositional (spectral) variation. The proposed approach can be used for exploratory visualization of MSI datasets without prior deep chemical or histological knowledge of the sample. Thus, different datasets can be visually compared employing the proposed method. The proposed approach allowed for the clear visualization of the molecular layer, granular layer, and white matter in chimpanzee and macaque cerebellum slices.

MSIWarp: A General Approach to Mass Alignment in Mass Spectrometry Imaging.

Mass spectrometry imaging (MSI) is a technique that provides comprehensive molecular information with high spatial resolution from tissue. Today, there is a strong push toward sharing data sets through public repositories in many research fields where MSI is commonly applied; yet, there is no standardized protocol for analyzing these data sets in a reproducible manner. Shifts in the mass-to-charge ratio (m/z) of molecular peaks present a major obstacle that can make it impossible to distinguish one compound from another. Here, we present a label-free m/z alignment approach that is compatible with multiple instrument types and makes no assumptions on the sample’s molecular composition. Our approach, MSIWarp (https://github.com/horvatovichlab/MSIWarp), finds an m/z recalibration function by maximizing a similarity score that considers both the intensity and m/z position of peaks matched between two spectra. MSIWarp requires only centroid spectra to find the recalibration function and is thereby readily applicable to almost any MSI data set. To deal with particularly misaligned or peak-sparse spectra, we provide an option to detect and exclude spurious peak matches with a tailored random sample consensus (RANSAC) procedure. We evaluate our approach with four publicly available data sets from both time-of-flight (TOF) and Orbitrap instruments and demonstrate up to 88% improvement in m/z alignment.

Understanding mass spectrometry images: complexity to clarity with machine learning

The application of artificial intelligence and machine learning to hyperspectral mass spectrometry imaging (MSI) data has received considerable attention over recent years. Various methodologies have shown great promise in their ability to handle the complexity and size of MSI data sets. Advances in this area have been particularly appealing for MSI of biological samples, which typically produce highly complicated data with often subtle relationships between features. There are many different machine learning approaches that have been applied to MSI data over the past two decades. In this review, we focus on a subset of non-linear machine learning techniques that have mostly only been applied in the past 5 years. Specifically, we review the use of the self-organizing map (SOM), SOM with relational perspective mapping (SOM-RPM), t-distributed stochastic neighbor embedding (t-SNE) and uniform manifold approximation and projection (UMAP). While not their only functionality, we have grouped these techniques based on their ability to produce what we refer to as similarity maps. Similarity maps are color representations of hyperspectral data, in which spectral similarity between pixels-that is, their distance in high-dimensional space-is represented by relative color similarity. In discussing these techniques, we describe, briefly, their associated algorithms and functionalities, and also outline applications in MSI research with a strong focus on biological sample types. The aim of this review is therefore to introduce this relatively recent paradigm for visualizing and exploring hyperspectral MSI, while also providing a comparison between each technique discussed.

Drosophila melanogaster chemical ecology revisited: 2-D distribution maps of sex pheromones on whole virgin and mated flies by mass spectrometry imaging

BackgroundSexual behavior in Drosophila melanogaster flies is greatly influenced by chemical cues. In this study, a spatial distribution of female and male sex pheromones was investigated on the surface of virgin and mated six-day-old fruit flies. Surface analysis was performed using the technique of mass spectrometry imaging (MSI) matrix-assisted laser desorption/ionization – time of flight (MALDI-TOF) and confirmed by gas chromatography – mass spectrometry (GC-MS) analysis of hexane extracts prepared from dissected flies.ResultsMALDI-TOF MSI experiments focused on female pheromones (7Z,11Z)-heptacosa-7,11-diene (7,11-HD) and (7Z,11Z)-nonacosa-7,11-diene (7,11-ND) were enhanced by using lithium 2,5-dihydroxybenzoate (LiDHB) matrix to improve the ionization and quality of MS spectra. Oxygenated compounds represented by male anti-attractants 11-cis-vaccenyl acetate (cVA) and 3-O-acetyl-1,3-dihydroxy-octacosa-11,19-diene (CH503) were successfully ionized without MALDI matrix in the form of potassium adducts in laser desorption ionization (LDI-TOF MSI) mode. A similar pattern of distribution for 7,11-HD and 7,11-ND was observed on the surface of both the left and the right female wing, with the strongest signals at the base of the wing. 7,11-HD and 7,11-ND were additionally detected on female legs but not on the body. The distribution of both male pheromones, cVA and CH503, was localized in virgin male flies solely on the abdominal tip and anogenital region. In addition, results from MSI experiments with mated flies showed the distribution of cVA and CH503 also on the female abdomen and thorax, demonstrating that anti-attractants were transferred from males to females during copulation. Results from LDI/MALDI-TOF MSI were supported by GC-MS analysis of hexane extracts prepared from different parts of virgin male or female Drosophila flies. Similar amounts of 7,11-HD and 7,11-ND were present on the legs, body and wings (127 ± 5 ng and 170 ± 8 ng, respectively). cVA was detected only on the male body. All acquired MSI datasets were affected by mass shift (predominantly between ±0.2 Da to ±0.4 Da), which was reduced using a mass recalibration approach.ConclusionsThe LDI/MALDI-TOF MSI technique makes it possible to study the distribution of female and male sexual pheromones on D. melanogaster flies. Moreover, the technique enables the transfer of male sex pheromones to females during copulation to be visualized. However, imaging experiments of 3-D biological samples performed on a single TOF-MS instrument equipped with a MALDI ion source and UV nitrogen laser evinced a photo-electric charging/discharging, a phenomenon that often leads to unpredictable mass shifts and poor mass accuracy.

Evaluation of Data Analysis Platforms and Compatibility with MALDI-TOF Imaging Mass Spectrometry Data Sets.

Imaging mass spectrometry (IMS) has proven to be a useful tool when investigating the spatial distributions of metabolites and proteins in a biological system. One of the biggest advantages of IMS is the ability to maintain the 3D chemical composition of a sample and analyze it in a label-free manner. However, acquiring the spatial information leads to an increase in data size. Due to the increased availability of commercial mass spectrometers capable of IMS, there has been an exciting development of different statistical tools that can help decipher the spatial relevance of an analyte in a biological sample. To address this need, software packages like SCiLS and the open source R package Cardinal have been designed to perform unbiased spectral grouping based on the similarity of spectra in an IMS data set. In this note, we evaluate SCiLS and Cardinal compatibility with MALDI-TOF IMS data sets of the Gram-negative pathogen Pseudomonas aeruginosa PA14. Both software were able to perform unsupervised segmentation with similar performance. There were a few notable differences which are discussed related to the identification of statistically significant features which required optimization of preprocessing steps, region of interest, and manual analysis.

Localized Metabolomic Gradients in Patient-Derived Xenograft Models of Glioblastoma.

Glioblastoma (GBM) is increasingly recognized as a disease involving dysfunctional cellular metabolism. GBMs are known to be complex heterogeneous systems containing multiple distinct cell populations and are supported by an aberrant network of blood vessels. A better understanding of GBM metabolism, its variation with respect to the tumor microenvironment, and resulting regional changes in chemical composition is required. This may shed light on the observed heterogeneous drug distribution, which cannot be fully described by limited or uneven disruption of the blood-brain barrier. In this work, we used mass spectrometry imaging (MSI) to map metabolites and lipids in patient-derived xenograft models of GBM. A data analysis workflow revealed that distinctive spectral signatures were detected from different regions of the intracranial tumor model. A series of long-chain acylcarnitines were identified and detected with increased intensity at the tumor edge. A 3D MSI dataset demonstrated that these molecules were observed throughout the entire tumor/normal interface and were not confined to a single plane. mRNA sequencing demonstrated that hallmark genes related to fatty acid metabolism were highly expressed in samples with higher acylcarnitine content. These data suggest that cells in the core and the edge of the tumor undergo different fatty acid metabolism, resulting in different chemical environments within the tumor. This may influence drug distribution through changes in tissue drug affinity or transport and constitute an important consideration for therapeutic strategies in the treatment of GBM. SIGNIFICANCE: GBM tumors exhibit a metabolic gradient that should be taken into consideration when designing therapeutic strategies for treatment.See related commentary by Tan and Weljie, p. 1231.

Analysis of Microbial Siderophores by Mass Spectrometry.

Siderophores represent important microbial virulence factors and infection biomarkers. Their monitoring in fermentation broths, bodily fluids, and tissues should be reproducible. Similar isolation, characterization, and quantitation studies can often have conflicting results, and without proper documentation of sample collection, data processing, and analysis methods, it is difficult to reexamine the data and reconcile these differences. In this Springer Nature Protocol, we present the procedure optimized for ferricrocin/triacetylfusarinine C extraction from biological material as well as for tissue fixation and cryosectioning for optical microscopy and for both elemental and molecular mass spectrometry imaging. Special attention is paid to siderophore data mining from conventional and product ion mass spectra, liquid chromatography, and mass spectrometry imaging datasets, performed here by our free software called CycloBranch.

Targeted Feature Extraction in MALDI Mass Spectrometry Imaging to Discriminate Proteomic Profiles of Breast and Ovarian Cancer.

To develop a mass spectrometry imaging (MSI) based workflow for extracting m/z values related to putative protein biomarkers and using these for reliable tumor classification. Given a list of putative breast and ovarian cancer biomarker proteins, a set of related m/z values are extracted from heterogeneous MSI datasets derived from formalin-fixed paraffin-embedded tissue material. Based on these features, a linear discriminant analysis classification model is trained to discriminate the two tumor types. It is shown that the discriminative power of classification models based on the extracted features is increased compared to the automatic training approach, especially when classifiers are applied to spectral data acquired under different conditions (instrument, preparation, laboratory). Robust classification models not confounded by technical variation between MSI measurements are obtained. This supports the assumption that the classification of the respective tumor types is based on biological rather than technical differences, and that the selected features are related to the proteomic profiles of the tumor types under consideration.

Data-Driven Rescoring of Metabolite Annotations Significantly Improves Sensitivity.

When analyzing mass spectrometry imaging data sets, assigning a molecule to each of the thousands of generated images is a very complex task. Recent efforts have taken lessons from (tandem) mass spectrometry proteomics and applied them to imaging mass spectrometry metabolomics, with good results. Our goal is to go a step further in this direction and apply a well established, data-driven method to improve the results obtained from an annotation engine. By using a data-driven rescoring strategy, we are able to consistently improve the sensitivity of the annotation engine while maintaining control of statistics like estimated rate of false discoveries. All the code necessary to run a search and extract the additional features can be found at https://github.com/anasilviacs/sm-engine and to rescore the results from a search in https://github.com/anasilviacs/rescore-metabolites .

Co-registration and analysis of multiple imaging mass spectrometry datasets targeting different analytes.

MALDI imaging mass spectrometry (IMS) has been successfully used to image a variety of biomolecules. Imaging of the many classes of biomolecules is often achieved through several incompatible sample preparations. Thus, multiple datasets must be acquired from multiple tissue sections to obtain a total molecular overview of a single sample. Addressing the need for single datasets from multiple IMS analyses, we developed the R package RegCombIMS as an extension of R package Cardinal to co-register, combine and create single IMS datasets acquired from serial sections of tissue. Dataset recombination and analysis is achieved by registration of the IMS datasets to a single coordinate space. The workflow allows for correlation of ions from IMS acquisitions that require incompatible sample preparations as well as multivariate analysis to mine the combined dataset for rapid and more thorough molecular query. The source code and example data are freely available at https://github.com/NHPatterson/RegCombIMS. All code was implemented in R. Supplementary data are available at Bioinformatics online.

Rapid discrimination of pediatric brain tumors by mass spectrometry imaging

PurposeMedulloblastoma, the most common primary pediatric malignant brain tumor, originates in the posterior fossa of the brain. Pineoblastoma, which originates within the pineal gland, is a rarer malignancy that also presents in the pediatric population. Medulloblastoma and pineoblastoma exhibit overlapping clinical features and have similar histopathological characteristics. Histopathological similarities confound rapid diagnoses of these two tumor types. We have conducted a pilot feasibility study analyzing the molecular profile of archived frozen human tumor specimens using mass spectrometry imaging (MSI) to identify potential biomarkers capable of classifying and distinguishing between medulloblastoma and pineoblastoma.MethodsWe performed matrix-assisted laser desorption ionization Fourier transform ion cyclotron resonance mass spectrometry imaging on eight medulloblastoma biopsy specimens and three pineoblastoma biopsy specimens. Multivariate statistical analyses were performed on the MSI dataset to generate classifiers that distinguish the two tumor types. Lastly, the molecules that were discriminative of tumor type were queried against the Lipid Maps database and identified.ResultsIn this pilot study we show that medulloblastoma and pineoblastoma can be discriminated using molecular profiles determined by MSI. The highest-ranking discriminating classifiers of medulloblastoma and pineoblastoma were glycerophosphoglycerols and sphingolipids, respectively.ConclusionWe demonstrate proof-of-concept that medulloblastoma and pineoblastoma can be rapidly distinguished by using MSI lipid profiles. We identified biomarker candidates capable of distinguishing these two histopathologically similar tumor types. This work expands the current molecular knowledge of medulloblastoma and pineoblastoma by characterizing their lipidomic profiles, which may be useful for developing novel diagnostic, prognostic and therapeutic strategies.

SPUTNIK: an R package for filtering of spatially related peaks in mass spectrometry imaging data.

Summary SPUTNIK is an R package consisting of a series of tools to filter mass spectrometry imaging peaks characterized by a noisy or unlikely spatial distribution. SPUTNIK can produce mass spectrometry imaging datasets characterized by a smaller but more informative set of peaks, reduce the complexity of subsequent multi-variate analysis and increase the interpretability of the statistical results.Availability and implementation SPUTNIK is freely available online from CRAN repository and at https://github.com/paoloinglese/SPUTNIK. The package is distributed under the GNU General Public License version 3 and is accompanied by example files and data.Supplementary information Supplementary data are available at Bioinformatics online.

Spatial Systems Lipidomics Reveals Nonalcoholic Fatty Liver Disease Heterogeneity.

Hepatocellular lipid accumulation characterizes nonalcoholic fatty liver disease (NAFLD). However, the types of lipids associated with disease progression are debated, as is the impact of their localization. Traditional lipidomics analysis using liver homogenates or plasma dilutes and averages lipid concentrations, and does not provide spatial information about lipid distribution. We aimed to characterize the distribution of specific lipid species related to NAFLD severity by performing label-free molecular analysis by mass spectrometry imaging (MSI). Fresh frozen liver biopsies from obese subjects undergoing bariatric surgery (n = 23) with various degrees of NAFLD were cryosectioned and analyzed by matrix-assisted laser desorption/ionization (MALDI)-MSI. Molecular identification was verified by tandem MS. Tissue sections were histopathologically stained, annotated according to the Kleiner classification, and coregistered with the MSI data set. Lipid pathway analysis was performed and linked to local proteome networks. Spatially resolved lipid profiles showed pronounced differences between nonsteatotic and steatotic tissues. Lipid identification and network analyses revealed phosphatidylinositols and arachidonic acid metabolism in nonsteatotic regions, whereas low–density lipoprotein (LDL) and very low–density lipoprotein (VLDL) metabolism was associated with steatotic tissue. Supervised and unsupervised discriminant analysis using lipid based classifiers outperformed simulated analysis of liver tissue homogenates in predicting steatosis severity. We conclude that lipid composition of steatotic and nonsteatotic tissue is highly distinct, implying that spatial context is important for understanding the mechanisms of lipid accumulation in NAFLD. MSI combined with principal component–linear discriminant analysis linking lipid and protein pathways represents a novel tool enabling detailed, comprehensive studies of the heterogeneity of NAFLD.