MS2Mol: A transformer model for illuminating dark chemical space from mass spectra

Butler, Thomas; Frandsen, Abraham; Lightheart, Rose; Bargh, Brian; Kerby, Thomas; West, Kiana; Davison, Joseph; Taylor, James; Krettler, Christoph; Bollerman, TJ; Voronov, Gennady; Moon, Kevin; Kind, Tobias; Dorrestein, Pieter; Allen, August; Colluru, Viswa; Healey, David

doi:10.26434/chemrxiv-2023-vsmpx-v4

Cited by 4 publications

(1 citation statement)

References 40 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…If prediction of target engagement becomes accurate and if experimental validation becomes widespread (these are big ifs), it follows that automation could accelerate the identification of potent and selective chemical matter, including natural products . The weak link here is procurement of materials through isolation , and synthesis.…”

Section: Discussionmentioning

confidence: 99%

Natural Product Synthesis in the 21st Century: Beyond the Mountain Top

Shenvi

2024

ACS Cent. Sci.

View full text Add to dashboard Cite

Research into natural products emerged from humanity's curiosity about the nature of matter and its role in the materia medica of diverse civilizations. Plants and fungi, in particular, supplied materials that altered behavior, perception, and well-being profoundly. Many active principles remain well-known today: strychnine, morphine, psilocybin, ephedrine. The potential to circumvent the constraints of natural supply and explore the properties of these materials led to the field of natural product synthesis. This research delivered new molecules with new properties, but also led to fundamental insights into the chemistry of the nonmetal elements H, C, N, O, P, S, Se, and their combinations, i.e., organic chemistry. It also led to a potent culture focused on bigger molecules and races to the finish line, perhaps at the expense of actionable next steps. About 20 years ago, the field began to contract in the United States. Research that focused solely on chemical reaction development, especially catalysis, filled the void. After all, new reactions and mechanistic insight could be immediately implemented by the chemistry community, so it became hard to justify the lengthy procurement of a complex molecule that sat in the freezer unused. This shift coincided with a divestment of natural product portfolios by pharmaceutical companies and an emphasis in academic organic chemistry on applications-driven research, perhaps at the expense of more fundamental science. However, as bioassays and the tools of chemical biology become widespread, synthesis finds a new and powerful ally that allows us to better deliver on the premise of the field. And the hard-won insights of complex synthesis can be better encoded digitally, mined by data science, and applied to new challenges, as chemists perturb and even surpass the properties of complex natural products. The 21st century promises powerful developments, both in fundamental organic chemistry and at the interface of synthesis and biology, if the community of scientists fosters its growth. This essay tries to contextualize natural product synthesis for a broad audience, looks ahead to its transformation in the coming years, and expects the future to be bright.

show abstract

Section: Discussionmentioning

confidence: 99%

Natural Product Synthesis in the 21st Century: Beyond the Mountain Top

Shenvi

2024

ACS Cent. Sci.

View full text Add to dashboard Cite

show abstract

Critical review on in silico methods for structural annotation of chemicals detected with LC/HRMS non-targeted screening

Hupatz,

Rahu,

Wang

et al. 2024

Anal Bioanal Chem

View full text Add to dashboard Cite

Non-targeted screening with liquid chromatography coupled to high-resolution mass spectrometry (LC/HRMS) is increasingly leveraging in silico methods, including machine learning, to obtain candidate structures for structural annotation of LC/HRMS features and their further prioritization. Candidate structures are commonly retrieved based on the tandem mass spectral information either from spectral or structural databases; however, the vast majority of the detected LC/HRMS features remain unannotated, constituting what we refer to as a part of the unknown chemical space. Recently, the exploration of this chemical space has become accessible through generative models. Furthermore, the evaluation of the candidate structures benefits from the complementary empirical analytical information such as retention time, collision cross section values, and ionization type. In this critical review, we provide an overview of the current approaches for retrieving and prioritizing candidate structures. These approaches come with their own set of advantages and limitations, as we showcase in the example of structural annotation of ten known and ten unknown LC/HRMS features. We emphasize that these limitations stem from both experimental and computational considerations. Finally, we highlight three key considerations for the future development of in silico methods. Graphical Abstract

show abstract

Pre-trained Maldi Transformers improve MALDI-TOF MS-based prediction

De Waele,

Menschaert,

Vandamme

et al. 2024

Preprint

View full text Add to dashboard Cite

For the last decade, matrix-assisted laser desportion/ionization time-of-flight mass spectrometry (MALDI-TOF MS) has been the reference method for species identification in clinical microbiology. Hampered by a historical lack of open data, machine learning research towards models specifically adapted to MALDI-TOF MS remains in its infancy. Given the growing complexity of available datasets (such as large-scale antimicrobial resistance prediction), a need for models that (1) are specifically designed for MALDI-TOF MS data, and (2) have high representational capacity, presents itself. Here, we introduce Maldi Transformer, an adaptation of the state-of-the-art transformer architecture to the MALDI-TOF mass spectral domain. We propose the first self-supervised pre-training technique adapted to mass spectra. The technique is based on shuffling peaks across spectra, and pre-training the transformer as a peak discriminator. Extensive benchmarks confirm the efficacy of this novel design. The final result is a model exhibiting state-of-the-art (or competitive) performance on downstream prediction tasks. In addition, we show that Maldi Transformer's identification of noisy spectra may be leveraged towards higher predictive performance. All code supporting this study is distributed on PyPI and is packaged under: https://github.com/gdewael/maldi-nn

show abstract

MS2Mol: A transformer model for illuminating dark chemical space from mass spectra

Cited by 4 publications

References 40 publications

Natural Product Synthesis in the 21st Century: Beyond the Mountain Top

Natural Product Synthesis in the 21st Century: Beyond the Mountain Top

Critical review on in silico methods for structural annotation of chemicals detected with LC/HRMS non-targeted screening

Pre-trained Maldi Transformers improve MALDI-TOF MS-based prediction

Contact Info

Product

Resources

About