Oscar Yanes

ChemEmbed: a deep learning framework for metabolite identification using enhanced MS/MS data and multidimensional molecular embeddings Abstract. Machine learning offers a promising path to annotating the large number of unidentified MS/MS spectra in metabolomics, addressing the limited cov

ChemEmbed: a deep learning framework for metabolite identification using enhanced MS/MS data and multidimensional molecular embeddings url: academic.oup.com/bib/article/...

A Perspective on Unintentional Fragments and their Impact on the Dark Metabolome, Untargeted Profiling, Molecular Networking, Public Data, and Repository Scale Analysis. In/post-source fragments (ISFs) arise during electrospray ionization or ion transfer in mass spectrometry when molecular bonds break, generating ions that can complicate data interpretation. Although ISFs have been recognized for decades, their contribution to untargeted metabolomics - particularly in the context of the so-called “dark matter” (unannotated MS or MS/MS spectra) and the “dark metabolome” (unannotated molecules) - remains unsettled. This ongoing debate reflects a central tension: while some caution against overinterpreting unidentified signals lacking biological evidence, others argue that dismissing them too quickly risks overlooking genuine molecular discoveries. These discussions also raise a deeper question: what exactly should be considered part of the metabolome? As metabolomics advances toward large-scale data mining and high-throughput computational analysis, resolving these conceptual and methodological ambiguities has become essential. In this perspective, we propose a refined definition of the “dark metabolome” and present a systematic overview of ISFs and related ion forms, including adducts and multimers. We examine their impact on metabolite annotation, experimental design, statistical analysis, computational workflows, and repository-scale data mining. Finally, we provide practical recommendations - including a set of dos and don’ts for researchers and reviewers - and discuss the broader implications of ISFs for how the field explores unknown molecular space. By embracing a more nuanced understanding of ISFs, metabolomics can achieve greater rigor, reduce misinterpretation, and unlock new opportunities for discovery.

If you’ve been following #metabolomics literature, you’ve probably seen a lot of debate on in-source fragmentation. We’ve put together a manuscript to clarify what it is, how to deal with it, and what it means for discovery in #metabolomics and #exposomics.
doi.org/10.26434/che...

SingleFrag: a deep learning tool for MS/MS fragment and spectral prediction and metabolite annotation Abstract. Metabolite and small molecule identification via tandem mass spectrometry (MS/MS) involves matching experimental spectra with prerecorded spectra

New paper out in Briefings in Bioinformatics

📰SingleFrag: a deep learning tool for MS/MS fragment and spectral prediction and metabolite annotation academic.oup.com/bib/article/...

7/ Compared to spray-coating, LTE shows:
✔️ Improved ionization
✔️ Reduced analyte diffusion
✔️ Better image sharpness
✔️ Cleaner baseline

8/ LTE pushes the boundaries of matrix deposition for MALDI-MSI. Better control, better data, better images. #MALDI #MassSpec #MSI #Metabolomics #Lipidomics

5/ LTE purifies the matrix during deposition — great for improving signal-to-noise, even when starting with lower-purity matrix

6/ The matrix stays stable at −80°C for at least 2 weeks. No loss in ionization efficiency or image quality after storage — big win for throughput & experimental planning

3/ We validated LTE using two matrices:
✅ DHB
✅ DAN
Calibrated thickness vs. deposition time = ✔️ reproducibility.

4/ ESEM images showed beautiful, uniform sub-micron matrix crystals across the tissue. Small crystals = better ionization = sharper images.

2/ We adapted Low-Temperature Thermal Evaporation (LTE)—originally used in nanotechnology and solar cell applications—for matrix deposition in #MALDI-MSI. The result: a reproducible, vacuum-based method that offers precise control over matrix thickness and produces ultra-pure coatings.

1/ In MALDI-MSI, matrix deposition is everything. It impacts sensitivity, spatial resolution, and reproducibility. We asked: can we improve matrix application using a dry, solvent-free, controlled method?

Improving MALDI Mass Spectrometry Imaging Performance: Low-Temperature Thermal Evaporation for Controlled Matrix Deposition and Improved Image Quality The deposition of matrix compounds significantly influences the effectiveness of matrix-assisted laser desorption/ionization (MALDI) Mass Spectrometry Imaging (MSI) experiments, impacting sensitivity, spatial resolution, and reproducibility. Dry deposition methods offer advantages by producing homogeneous matrix layers and minimizing analyte delocalization without the use of solvents. However, refining these techniques to precisely control matrix thickness, minimize heating temperatures, and ensure high-purity matrix layers is crucial for optimizing MALDI-MSI performance. Here, we present a novel approach utilizing low-temperature thermal evaporation (LTE) for organic matrix deposition under reduced vacuum pressure. Our method allows for reproducible control of matrix layer thickness, as demonstrated by linear calibration for two organic matrices, 2,5-dihydroxybenzoic acid (DHB) and 1,5-diaminonaphthalene (DAN). The environmental scanning electron microscopy images reveal a uniform distribution of small-sized matrix crystals, consistently on the sub-micrometer scale, across tissue slides following LTE deposition. Remarkably, LTE serves as an additional purification step for organic matrices, producing very pure layers irrespective of initial matrix purity. Furthermore, stability assessment of MALDI-MSI data from mouse brain sections coated with LTE-deposited DHB or DAN matrix indicates minimal impact on ionization efficiency, signal intensity, and image quality even after storage at −80 °C for 2 weeks, underscoring the robustness of LTE-deposited matrices for MSI applications. Comparative analysis with the spray-coating method highlights several advantages of LTE deposition, including enhanced ionization, reduced analyte diffusion, and improved MSI image quality.

We’re excited to share our latest work in @jasms.bsky.social:
“Improving MALDI Mass Spectrometry Imaging Performance: Low-Temperature Thermal Evaporation for Controlled Matrix Deposition and Improved Image Quality”
🧵👇 pubs.acs.org/doi/10.1021/...

I created a Metabolomics starter pack. A list of researchers from the wonderful world of #metabolomics. If you would like to be added (or removed) just let me know. go.bsky.app/J3VPYKm

Delighted to put our grain of sand into this fascinating work with @manelesteller.bsky.social #aging #metabolism #metabolomics #omics

ONLY 3 DAYS LEFT for computational biologists ...!

To apply for our PhD student position
- in Metabolic Genome Regulation
- in aAML
- cosupervised by Tanya Vavouri and me
- at Josep Carreras Institute
- in Barcelona, Spain
- embedded into the MSCA HubMOL network

Apply here: hubmol.eu

Please RP!

ChemEmbed: A deep learning framework for metabolite identification using enhanced MS/MS data and multidimensional molecular embeddings Machine learning tools have become essential for annotating the vast number of unidentified MS/MS spectra in metabolomics, addressing the limitations of current reference spectral libraries. However, ...

5/ Want to learn more? 📄
Read our full paper on #bioRxiv: www.biorxiv.org/content/10.1...
#Metabolomics #MachineLearning #DeepLearning #MSMS
We’d love to hear your thoughts!
This is another successful collaboration with @seeslab.bsky.social at @urv.cat

4/ The results:
✅ ChemEmbed ranks the correct metabolite #1 in 42% of cases in a test dataset.
✅ Finds the correct compound in the top 5 in 76% of cases
✅ Against external benchmarks CASMI 2016 and 2022, and ARUS dataset (unidentified spectra from human plasma & urine), ChemEmbed outperforms #SIRIUS

3/ We enhance MS/MS data by:
✅ Merging spectra from multiple collision energies
✅ Incorporating calculated neutral losses
✅ Training a CNN on a dataset of 38,472 unique compounds from NIST20, MSDIAL, GNPS, and Agilent METLIN metabolomic libraries

2/ Our solution to reduce this problem: #ChemEmbed
We combine enhanced MS/MS spectra with continuous vector representations of molecular structures (300-dimensional embeddings aligned with Mol2vec representations). This gives our CNN-based model richer input, improving annotation accuracy.

1/ The problem:
#Metabolomics relies on MS/MS spectral databases, but most spectra remain unidentified due to limited reference libraries. Computational methods help, but they struggle with high-dimensional and sparse spectral and structural data.

🚀 New paper alert! 🚀
Happy to introduce #ChemEmbed, a deep learning framework for metabolite identification that enhances MS/MS data and leverages multidimensional molecular embeddings. A 🧵 on how it works and why it matters! ⬇️ #metabolomics #MachineLearning #DeepLearning

📢📢 Oferta de Feina a MIL@b 📢 📢
- Posició: Tècnic de Recerca en Metabolòmica
- Ubicació: Tarragona
- Data límit d'inscripció: 17 de gener de 2025
- Inscripció i aplicacions: A través de la seu oficial de la URV (tinyurl.com/muhaxvhv)
- Més informació 👇:

📢📢 OPEN POSITION at MIL@b📢 📢
Position: Research Scientist - LCMS, GC/MS specialist for metabolomics
Placement: Tarragona, Spain
Registration deadline: 2025, January 17th
Applications: Follow URV official linktinyurl.com/muhaxvhv
Further information 👇:

Improving MALDI Mass Spectrometry Imaging Performance: Low-Temperature Thermal Evaporation for Controlled Matrix Deposition and Improved Image Quality The deposition of matrix compounds significantly influences the effectiveness of matrix-assisted laser desorption/ionization (MALDI) Mass Spectrometry Imaging (MSI) experiments, impacting sensitivity,...

Sorry, now it should work www.biorxiv.org/content/10.1...

Improving MALDI Mass Spectrometry Imaging Performance: Low-Temperature Thermal Evaporation for Controlled Matrix Deposition and Improved Image Quality The deposition of matrix compounds significantly influences the effectiveness of matrix-assisted laser desorption/ionization (MALDI) Mass Spectrometry Imaging (MSI) experiments, impacting sensitivity,...

www.biorxiv.org/content/10.1...

An ongoing study in our lab focuses on developing new matrices for MALDI-MSI of small molecules (m/z <400) using LTE deposition. We're minimizing matrix background interference while enhancing metabolite ionization and coverage, paving the way for #Spatial #Metabolomics. More exciting results soon!

🔬 Read more about how LTE advances the field of MALDI-MSI, including crystal morphology, stability tests, and comparative analysis with spray-coating, in our new preprint: biorxiv.org/content/10.110… #MALDIMS #MassSpectrometry #Lipidomics

Spray-coating is the most popular method, but LTE outperforms it by:
🔹 Enhancing ionization efficiency
🔹 Reducing analyte diffusion
🔹 Improving spatial resolution & image quality in MSI images

Stability is another win! After storing mouse brain sections coated with DHB or DAN matrices at -80 °C for two weeks, ionization efficiency, signal intensity, and image quality remained consistent. Robustness is crucial for long-term studies, and LTE delivers.

Our LTE deposition method works under reduced vacuum pressure, enabling precise control over matrix thickness. With linear calibrations for DHB & DAN, LTE ensures a reproducible matrix layer thickness—critical for consistent results!

💡 Why focus on matrix deposition? It’s a key factor in MALDI-MSI experiments, influencing data quality, analyte localization, and experimental reproducibility. Traditional methods have limitations, so we aimed to improve the game.

My first thread in #Bluesky, let's go then: Excited to share our latest study on a novel approach for #MALDI-#MSI matrix deposition! We’ve developed a low-temperature thermal evaporation (LTE) method that optimizes sensitivity, spatial resolution, and reproducibility. 🧵