Mass Spectrum Partial Structure Estimation Using Fragment Peaks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for estimating partial structures of unknown compounds from mass spectra lack precision, particularly in distinguishing fragment peaks from other signals, leading to degraded estimation accuracy.
Innovation Solution
A partial structure estimation apparatus and method that generate explanatory variables from peak and peak interval compositions using machine learning, focusing solely on fragment peaks to improve estimation precision by registering and generating models based on known compounds' mass spectra.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If all peaks in the mass spectrum are used for partial structure estimation, then the estimation can be performed using available data, but the precision is degraded due to interference from non-fragment peaks
Solution Approach 1:
The mass spectrum peaks are segmented into fragment peaks and non-fragment peaks based on their characteristics. The estimation model processes only the fragment peak data, separating the useful information from interfering signals to improve estimation precision.
Solution Approach 2:
The method extracts and selects only the fragment peaks from the complete set of mass spectrum peaks. By taking out the relevant fragment peak information and excluding non-fragment peaks, the model achieves more accurate partial structure estimation without interference from irrelevant signals.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A partial structure estimation apparatus (20) is configured to generate a first explanatory variable by performing composition estimation for each peak in a mass spectrum acquired from a sample, and to generate a second explanatory variable by performing composition estimation for each peak interval in the mass spectrum. The partial structure estimation apparatus (20) is further configured to then estimate a partial structure as an objective variable based on the first explanatory variable and the second explanatory variable. In a partial structure estimation model generation apparatus (12), a partial structure estimation model (18) is generated through machine learning using a training data set.