Neural Network Matrix Classifier for Multidimensional Mass Spectrometry

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Multidimensional mass spectrometry data analysis is complex due to the interdependence of components, requiring improved techniques for extracting specific insights from raw data.

Innovation Solution

A system and method utilizing a mass spectrometer, matrix generator, and matrix classifier, where the multidimensional data set is converted into a matrix or image, and a trained neural network is used to determine the class of the matrix, allowing for enhanced classification and analysis.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Loss of information

If traditional routines are used for extraction of specific information from raw mass spectrometry data, then specific component information can be obtained, but additional insights about interdependence between components are lost

Engineering Contradiction:

Improveinformation about interdependence between componentsVSAvoidcomplexity of data analysis technique

Core Design Contradiction:

Loss of informationVSDevice complexity

Solution Approach 1:

The patent merges multiple data dimensions (retention time, mass-to-charge ratio, ion intensity) into a unified matrix representation that preserves all component information and their interdependencies simultaneously, rather than extracting specific components separately through traditional routines

Inventive Principle:

Principle #5Merging (Combining)

Solution Approach 2:

The patent transforms multidimensional mass spectrometry data into a matrix format where retention time and mass-to-charge ratio form two dimensions, and ion intensity provides a third dimension, enabling comprehensive analysis of component interdependencies through the additional dimensional structure

Inventive Principle:

Principle #17Another dimension (Dimensionality change)

2Measurement precision

If multidimensional data sets are analyzed using traditional extraction routines, then processing is straightforward, but classification accuracy and insight extraction are limited

Engineering Contradiction:

Improveclassification accuracyVSAvoidcomplexity of analysis system

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent introduces a matrix generator as an intermediary component that converts raw multidimensional mass spectrometry data into a structured matrix representation, which then serves as input for the matrix classifier, improving classification accuracy while managing system complexity through modular design

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent changes the parameter representation by transforming raw data parameters (retention time, mass-to-charge ratio, ion intensity) into matrix parameters (matrix cell location, matrix cell value), enabling more effective classification through the transformed parameter space

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentEP3584795B13D mass spectrometry predictive classification

Publication Date: 2022.10.19 THERMO FINNIGAN LLC

Figure 1
Figure 2A~2B
Figure 3

AI summary

A method for analyzing a multidimensional data set includes generating a multidimensional mass spectrometry data set from a sample; and generating an matrix representing the multidimensional mass spectrometry data set such that a first dimension and a second dimension of the multidimensional mass spectrometry data set correspond to a matrix cell location, and an ion intensity corresponds to a matrix cell value; and determining a class of the matrix from a plurality of matrix classes using a trained neural network matrix classifier.