Mass Spectrometer ML Module for False Positive Elimination

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current mass spectrometers rely on commercial software for Tentatively Identified Compounds (TICs) selection, resulting in numerous false positives due to solely library-based matching, requiring extensive manual inspection and being prone to human error, which is time-consuming and inefficient.

Innovation Solution

Integration of a machine learning-based computer module into mass spectrometers to reduce false positives by using a cubic linear regression model, calculating scores for chemical identification accuracy, and automating the elimination process, thereby reducing the timeframe for eliminating false positives to a couple of hours.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If library-based matching software (NIST/AMDIS) is used for chemical identification, then identification capability is provided, but numerous false positives are generated requiring extensive manual inspection

Engineering Contradiction:
Improveidentification throughputVSAvoididentification accuracy
Core Design Contradiction:
ProductivityVSReliability

Solution Approach 1:

The patent introduces an intermediary machine learning model that acts as a mediator between the library matching software and final identification results. This ML model processes the initial matches and filters out false positives, reducing manual inspection needs while maintaining high accuracy. The intermediary layer reconciles the high throughput of library matching with the need for reliable identification.

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent replaces the manual mechanical inspection process with an automated machine learning-based evaluation system. Instead of human experts manually reviewing each match, the ML model automatically scores and ranks identifications, eliminating the time-consuming manual verification step while improving consistency and reliability.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

2Reliability

If manual inspection of tentative identified compounds is performed, then false positives can be eliminated, but considerable effort in days/weeks is required and human error may influence results

Engineering Contradiction:
Improvefalse positive eliminationVSAvoidinspection timeframe
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The patent substitutes the manual human inspection process with an automated machine learning evaluation system. The ML model processes identifications in minutes rather than days or weeks, eliminating false positives with high accuracy while removing human error and fatigue from the process. The automated system provides consistent, reproducible results without the time investment required for manual review.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The system enables self-service by allowing the machine learning model to autonomously evaluate and rank tentative identifications without human intervention. The ML model independently scores matches, identifies false positives, and provides confidence rankings, making the system self-sufficient and eliminating the need for time-consuming manual verification.

Inventive Principle:
Principle #25Self-service

3Extent of automation

If solely library match is used for compound identification, then automated identification is achieved, but numerous false positives result

Engineering Contradiction:
Improveautomated identificationVSAvoididentification accuracy
Core Design Contradiction:
Extent of automationVSReliability

Solution Approach 1:

The patent introduces a machine learning evaluation model as an intermediary layer between automated library matching and final identification results. This ML intermediary processes the automated matches, scores them based on multiple criteria including spectral quality and chemical plausibility, and filters out false positives, thereby maintaining automation while significantly improving identification reliability.

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent changes the evaluation parameters from simple library match scoring to a multi-parameter machine learning assessment that considers spectral quality, chemical structure plausibility, and confidence metrics. By transforming the identification criteria from single-parameter library matching to multi-parameter ML evaluation, the system maintains automation while dramatically reducing false positives.

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentUS20240071738A1Method of improving a mass spectrometer, module for improving a mass spectrometer and an improved mass spectrometer
Publication Date: 2024.02.29 THE GOVERNMENT OF THE UNITED STATES AS REPRESENTED BY THE SECRETARY OF THE AIR FORCE

AI summary

The present invention relates to a method of improving a mass spectrometer, a module for improving a mass spectrometer and an improved mass spectrometer. The aforementioned method uses machine learning and can greatly reduce the number and timeframe for the elimination of false positives to a couple of hours. Such method can be integrated into a mass spectrometer by inserting a computer module programmed with such method into a mass spectrometer's computer system.