Identification of metabolites and small compounds using mass spectrometry

质谱是鉴定小分子的关键技术。如今，从质谱中鉴定代谢物依赖于与真实化合物或参考谱的比较。第一种方法往往很昂贵，而后者只覆盖了巨大化学空间的一小部分。ChemFrag项目的主要目标是利用串联质谱法从通用化合物库中获得的候选化合物中鉴定化合物。由于该过程的机械模拟在计算上是不可行的，我们开发了简化的计算机碎片化方法，统计模型，并将机器学习应用于大量的训练光谱。我们的方法筛选了一组候选分子，以找到解释实验测得的峰的分子片段。除了一个简单的键断开的方法在分子图上，我们将化学背景和预先计算的键能和键应变质子化。

Mass spectrometry is a key technology for the identification of small molecules. Today, the identification of metabolites from mass spectra relies on the comparison with authentic compounds or reference spectra. The first are often expensive to obtain, the latter cover only a small portion of the vast chemical space. The key objective of the ChemFrag project is to identify compounds with tandem mass spectrometry in candidates obtained from general purpose compound libraries. Because a mechanistic simulation of the process is computationally infeasible, we develop simplified in-silico fragmentation methods, statistical models and apply machine learning to a large set of training spectra. Our approach screens a set of candidate molecules, to find molecular fragments which explain the experimentally measured peaks. In addition to a simple bond disconnection approach on the molecular graph, we incorporate the chemical context and precalculated bond energies and bond strain upon protonation.