![[2602.16507] Small molecule retrieval from tandem mass spectrometry: what are we optimizing for?](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.16507] Small molecule retrieval from tandem mass spectrometry: what are we optimizing for?
Summary
This paper explores the optimization of loss functions in deep learning models for small molecule retrieval from tandem mass spectrometry data, revealing a trade-off between fingerprint accuracy and molecular retrieval performance.
Why It Matters
Understanding the impact of loss functions on model performance is crucial for improving the accuracy of small molecule identification in mass spectrometry. This research provides insights that can guide the selection of loss functions and molecular fingerprints, enhancing the effectiveness of computational analysis in chemistry and related fields.
Key Takeaways
- Investigates the impact of various loss functions on model performance in mass spectrometry.
- Reveals a trade-off between fingerprint similarity and molecular retrieval accuracy.
- Provides novel regret bounds for understanding decision-making in model optimization.
- Guides the selection of loss functions based on the similarity structure of candidate sets.
- Highlights the importance of balancing accuracy in predictions with retrieval outcomes.
Computer Science > Machine Learning arXiv:2602.16507 (cs) [Submitted on 18 Feb 2026] Title:Small molecule retrieval from tandem mass spectrometry: what are we optimizing for? Authors:Gaetan De Waele, Marek Wydmuch, Krzysztof Dembczyński, Wojciech Kotłowski, Willem Waegeman View a PDF of the paper titled Small molecule retrieval from tandem mass spectrometry: what are we optimizing for?, by Gaetan De Waele and 4 other authors View PDF HTML (experimental) Abstract:One of the central challenges in the computational analysis of liquid chromatography-tandem mass spectrometry (LC-MS/MS) data is to identify the compounds underlying the output spectra. In recent years, this problem is increasingly tackled using deep learning methods. A common strategy involves predicting a molecular fingerprint vector from an input mass spectrum, which is then used to search for matches in a chemical compound database. While various loss functions are employed in training these predictive models, their impact on model performance remains poorly understood. In this study, we investigate commonly used loss functions, deriving novel regret bounds that characterize when Bayes-optimal decisions for these objectives must diverge. Our results reveal a fundamental trade-off between the two objectives of (1) fingerprint similarity and (2) molecular retrieval. Optimizing for more accurate fingerprint predictions typically worsens retrieval results, and vice versa. Our theoretical analysis shows this trade-o...