Resumen de Computational tools for the annotation of in-source fragments and matrix-related signals in MALDI Mass Spectrometry Imaging
- español
La espectrometría de masas de imagen MALDI (MALDI-MSI) es una técnica analítica utilizada en estudios bioquímicos y clínicos para revelar la composición química y la información espacial de tejidos orgánicos. Los fragmentos generados dentro de la fuente MALDI y las señales relacionadas con la matriz abarrotan los espectros, lo que hace que la identificación de cada masa a carga (m/z) sea un desafío. En esta tesis, desarrollamos dos herramientas computacionales (rMSIfragment y rMSIcleanup) para la anotación controlada por FDR de fragmentos en la fuente y señales relacionadas con la matriz. También presentamos un protocolo computacional y experimental completo basado en una matriz MALDI marcada con isótopos estables para descubrir señales relacionadas con la matriz. Demostramos el alto rendimiento de nuestras herramientas y protocolos en múltiples tipos de muestras, matrices MALDI y analizadores MS. También encontramos que la eliminación de estas señales permite que las técnicas de reducción de dimensionalidad se centren mejor en las características espaciales biológicamente relevantes y mejora la anotación de metabolitos. En conjunto, estos resultados indican que la anotación de fragmentos en la fuente y las señales relacionadas con la matriz deben incluirse en estudios rigurosos de metabolómica no dirigidos utilizando MSI.
- català
L´espectrometria de masses d´imatge MALDI (MALDI-MSI) és una tècnica analítica utilitzada en estudis bioquímics i clínics per revelar la composició química i la informació espacial de teixits orgànics. Els fragments generats dins de la font MALDI i els senyals relacionats amb la matriu abarroten els espectres, cosa que fa que la identificació de cada massa a càrrega (m/z) sigui un desafiament. En aquesta tesi, desenvolupem dues eines computacionals (rMSIfragment i rMSIcleanup) per a l'anotació controlada per FDR de fragments a la font i senyals relacionats amb la matriu. També presentem un protocol computacional i experimental complet basat en una matriu MALDI marcada amb isòtops estables per descobrir senyals relacionats amb la matriu. Demostrem l'alt rendiment de les nostres eines i protocols en múltiples tipus de mostres, matrius MALDI i analitzadors MS. També trobem que l'eliminació d'aquests senyals permet que les tècniques de reducció de dimensionalitat se centrin millor en les característiques espacials biològicament rellevants i millorin l'anotació de metabòlits. En conjunt, aquests resultats indiquen que l'anotació de fragments a la font i els senyals relacionats amb la matriu s'han d'incloure en estudis rigorosos de metabolòmica no dirigits utilitzant MSI.
- English
Justification and Needs for the Research: Matrix-Assisted Laser Desorption Ionization Mass Spectrometry Imaging (MALDI-MSI) is an analytical technique used in biochemical and clinical studies to reveal the chemical composition and spatial information of organic tissues. It provides valuable information in many applications, including the understanding and diagnosis of complex diseases such as cancer, diabetes, Alzheimer’s and infectious diseases.
Despite the surge of MALDI-MSI’s popularity, associating each mass-to-charge (m/z) signal with univocal molecular identifications remains challenging: (1) samples include thousands of molecules; (2) each molecule is responsible for several MS signals (isotopes, adducts, in-source fragments, multiple charges…); and (3) isomers and isobars cannot be resolved using only MS1.
Traditional mass spectrometry techniques rely on chromatographic separation (LC-MS, GC-MS) for sample simplification. However, MALDI-MSI does not include such separation steps. Complementary, tandem mass spectrometry can augment the depth of the chemical analysis by providing fragmentation information on single molecules. Many MALDI-MSI instruments are equipped with tandem-MS capabilities (Bruker’s ultrafleXtreme, Thermo Scientific’s MALDI LTQ Orbitrap XL, or Waters’ MALDI SYNAPT G2-Si) but untargeted imaging MS/MS is not routinely feasible due to (1) prohibitive running times, (2) limited parental ion selectivity and intensity, and (3) increased data size and complexity. For all these reasons, untargeted fragmentation of all ions in a sample is only possible using highly specialized instrumental setups.
In this complex scenario, the annotation of MS signals and putative identification of metabolites present in the sample is a daunting task. There are several software solutions to perform automatic annotation of MSI data. However, two types of signals have been traditionally overlooked and underestimated: in-source fragments and matrix-related signals.
In-Source Decay (ISD) or In-Source Fragmentation (ISF) (i.e. the natural and unavoidable generation of fragments inside the MALDI ion source) needs to be minimized. ISD depends mainly on the chemical structure of the analyte and ionization conditions such as ionization temperature or voltage and can be problematic in the study of lipids, as several fragmentation pathways lead to isobaric lipid species. These known lipid fragmentation pathways result in falsely low concentrations of lipids suffering from ISD and falsely high concentrations of lipids overlapping with isobaric in-source fragments. Additionally, if not properly annotated and removed, in-source fragments can yield an increased number of incorrect annotations using common MALDI-MSI annotation tools such as LipoStar MSI, METASPACE, and rMSIannotation.
On the other hand, in the classical MALDI-MSI workflow, an organic compound (e.g. matrix) is deposited onto the sample to promote the desorption and ionization of endogenous analytes. Unfortunately, this low-weight exogenous compound adds several undesired MS signals to the MALDI-MSI spectra. Including exogenous matrix signals (adducts, multiple charges, and in- source fragments) and matrix adducts with endogenous biomolecules. These signals add an undesired layer of complexity to core MSI processing pipelines like untargeted statistical analyses or molecular annotation. This is particularly worrying in metabolomics and lipidomics, as matrix-related signals are densely concentrated in the low m/z range.
Current automated annotation solutions for MSI overlook in-source fragments and matrix-related signals. There is a clear need to develop such tools.
Methodology: To address the automatic annotation of in-source fragments we develop rMSIfragment. An open-source R package that exploits known in-source fragmentation pathways to increase confidence in lipid annotations. Our novel ranking score combines the times a given lipid has been found in the dataset (adducts and in-source fragments) and their spatial correlation to filter out unlikely lipids.
We validate our tool on a wide range of samples acquired with MALDI-MSI. Including 15 human nevi samples and 12 publicly available datasets covering different sample types, preparations, and acquisition parameters. We match our annotations against HPLC as well as a new Target-Decoy algorithm that introduces decoy in-source fragmentation pathways.
To address the automatic annotation of matrix-related signals we first develop rMSIcleanup. An open-source R package that relies on the spectral similarity and spatial similarity of ions pertaining to the same molecule to annotate signals related to the matrix. Additionally, we include an algorithm for peak overlap detection making the package also suitable with lower resolving power MS analyzers.
We validate the package using 14 datasets acquired with Ag-LDI-MSI, where the traditional organic matrix is replaced by silver sputtering. In a later study we demonstrate its applicability to the annotation of DHB-related signal, the most widely used MALDI matrix. We use a total of 30 MSI datasets covering different sample types and MS analysers.
Finally, we acknowledge that current automatic tools for the annotation of matrix-annotation tools suffer from multiple of the following pitfalls: (1) focus exclusively on the spatial distribution, (2) do not control the False Discovery Rate (FDR), (3) do not consider adducts with endogenous metabolites, and (4) rely on a predefined list of theoretical matrix adducts.
To solve all of these issues we develop an experimental and computational workflow to discover matrix-containing adducts using 13C^6-labeled 2,5-Dihydroxybenzoic acid (13C^6-DHB). By exploiting the labeling-induced m/z shift and unique spatial distribution of matrix-containing ions we can discover and annotate matrix-containing adducts formed with exogenous and endogenous compounds. Key computational novelties include the registration of the technical replicates to enable correlation across samples and the definition of a decoy database based on a decoy matrix and decoy m/z shifts to estimate the FDR. Conclusions: A few main conclusions permeate from this thesis. Firstly, rMSIfragment can annotate in-source fragments in lipids. Interestingly we found that, if not properly annotated, in-source fragments of lipids can be overlapped with up to 50% of lipid parental annotations. rMSIfragment mitigates this issue by two mechanisms (1) the user is made aware of the overlap so they can be cautious with the interpretation of the results and (2) highly unlikely parental or fragment ions are deprioritized with a low-ranking score thus effectively reducing the number of overlaps.
Secondly, we demonstrated that rMSIcleanup can confidently annotate matrix-related signals, both in matrix-free approaches such as Ag-LDI-MSI and when using organic matrices such as DHB. We demonstrated that the proper annotation and removal of in-source fragments and matrix-related signal has a significant positive impact on downstream untargeted metabolomics workflow. As an example, the removal of matrix-related signals allows dimensionality reduction techniques such as PCA and UMAP to better focus on biologically relevant difference. We hypothesise this is because matrix-related signals introduce redundant and non-biologically relevant information. Perhaps more interesting is the realization that the removal of matrix-related signals also improves metabolite annotation when common tools such as METASPACE. We rationalize this improvement by interpreting how the Target and Decoy searches behave when the matrix-related signals are removed. While in the search of common Finally, we propose a new methodology based on SIL-MALDI-MSI to create a curated list of matrix-related signals. This allows us to demonstrate the high prevalence of DHB adducts with endogenous compounds and helps shed some light into the vast number of unannotated signals in the typical MSI experiment. This methodology can be applied to other matrices, tissue types and analysers.
