Trends in Computational Metabolomics: A Perspective on Five Years of Software Development, Challenges, and Opportunities (2021-2025).

Metabolomics software development has accelerated rapidly, yet no recent systematic analysis has quantified how the landscape is evolving across computational methods, geographies, and the research community's technology adoption. There is a strong need within the metabolomics research community to keep pace with the rapid expansion of accessible and free computational tools and resources. Given the absence of such a treatise since 2021 and the surge in advances in ion mobility mass spectrometry (IM-MS), single-cell and spatial metabolomics, and multimodal omics-based discovery, we offer a curated database that aggregates 746 mass spectrometry- and spectroscopy-based tools across 37 categories from data preprocessing to metabolite annotation. We report four structural shifts that redefine the field's trajectory. First, machine learning (ML) adoption in tools increased by 2.4-fold from 10.9% (2021) to 26.6% (2025). Second, annotation as a category commands the most tools (16.8%) and the highest ML investment among any of the proposed tool categories. The dominant strategy has shifted from library matching (2021) to spectrum prediction (2024) and, more recently, to de novo structure generation (2025), thereby progressively reducing the reliance on accessible experimental spectral reference databases. Third, Python has displaced R as the dominant programming language, with a sharp inflection in 2023 coinciding with the ML surge, while web server-only tools have sharply declined. Fourth, transformer architectures grew significantly, and in 2025, the first few large language model (LLM)-based and other multimodal metabolomics tools emerged, signaling a transition from task-specific classifiers toward pretrained, transferable representations. Concurrently, adoption of preprints as a publishing venue also rose by 2.5-fold, and, notably, mentions of benchmarking and explainability each increased by 8-18-fold, indicating a growing community-wide need and maturation. This computational metabolomics database is now made available here: https://github.com/enveda/computational-metabolomics-review.