Automating GC/MS Analysis for High-Quality Data

The Agilent AI Peak Integration for MassHunter Software enables researchers to automate their data analysis workflow and enhance lab productivity.

Q: What is the difference between the Agilent AI Peak Integration for MassHunter Software versus other automated parameter-less integrators?

A: GC/MS data analysis is often manual, time-consuming, and error-prone. Even though many automated integrators are available, a chemist may still need to manually correct integration peaks and baselines. The Agilent AI Peak Integration for MassHunter Software is a quantitative analysis software add-on that leverages machine learning (ML) to automate peak integration during the data analysis process for analytical testing laboratories. In contrast to other solutions, the Agilent software not only integrates the peak but keeps learning from your manual integration performance. The ML model is custom-trained through a user’s standardized data analysis workflow by observing manual peak integration events and replaces manual peak integration performed by a user with 1) adaptable artificial intelligence (AI)-assisted peak detection, and 2) continuous learning integration.

Q: What are the key benefits to analytical testing laboratories in using the Agilent AI Peak Integration for MassHunter Software?

A: The novel peak integration software helps maximize sample throughput, improve process automation, and deliver high-quality, consistent results while ensuring ease of adoption for all users throughout the lab. Effectively, a laboratory chemist now can maximize the time spent on different workflow activities. To summarize, the Agilent AI Peak Integration add-on offers reproducible and universally consistent analytical results with a reduced turnaround time that helps improve overall lab efficiency while ensuring ease of use for expert and new lab analysts.

Q: How does this software change the current workflow in analytical test laboratories?

A: There is minimal change to the current workflow at the analytical test laboratories. Agilent GC/MS instruments are initially utilized to acquire samples with the Agilent MassHunter Workstation software enabling data acquisition and quantitation. Agilent AI Peak Integration for MassHunter is an add-on for MassHunter Quant and helps communicate to the cloud-based AI Peak Integration database. Samples and methods are then combined in a MassHunter Batch that a chemist continues to analyze per the batch sequence. The added bonus here is that rather than the chemist performing manual integration on the data, the AI Peak Integration machine learning model can be used to automate processing further optimizing the overall workflow.

Q: What’s the percentage in terms of time saved when using the Agilent AI Peak Integration for MassHunter Software?

A: Under ideal conditions with an expert analyst, reviewing data and performing manual integration requires an average of 60-120 seconds per chromatogram. The AI Peak Integration solution outpaces manual integration after just a few positive samples. The minimum cloud processing time is about 30 seconds. In preliminary trials, AI Peak Integration software performed peak integration on 100 samples in fewer than 25 minutes, whereas an expert chemist (under the ideal conditions) would take 2 hours to manually integrate peaks in the same batch, thus enabling at least a 4-times gain in productivity.

Q: What are the target applications for the Agilent AI Peak Integration for MassHunter Software?

A: Machine Learning modeling holds promise for several applications in GC/MS. This product is currently only intended for Phthalates and Tris (1,3-dichloro- 2-propyl) phosphate (“TDCPP”) analysis due to the complexity of peak integrations for these compounds. This peak integration solution will be expanded to other applications and markets in the future. Environmental semi-volatile organic compounds are candidates for future experimentation.

Q: Why select Phthalates as the target application in your first release?

A: Phthalate compounds are heavily used in packaging and plastic materials. They present several challenges for both LC/MS and GC/MS. To start, fragmentation patterns are extremely similar and suffer mass spectrometric identification. High molecular weight phthalates exist as a mixture of isomers which impacts chromatography and subsequent manual integration of the compounds’ high molecular weight proves difficult, even for an experienced analyst.

Q: How can we ensure the performance of the model?

A: Iterative optimization techniques are applied to enhance the performance of the AI model. This involves establishing the optimal model training regime, cleaning the dataset, handling the outliers, and incorporating feedback from domain experts. Performance metrics, such as accuracy and correctness metrics, are utilized to constantly monitor and improve the model. A comparison was made between the AI model’s performance, the MassHunter default parameter-less integrator, and manual integrations, performed by scientists to establish the optimal in-production prediction mode. How the machine learning model is trained is a key step; Agilent application engineers will provide customers assistance with ML model training during the initial installation. Additionally, customers have access to a model performance and monitoring portal where integration performance can be compared to the customers’ original manual integration performance. This aids in evaluating the precision and sensitivity accuracy with AI-Integrated MH.

Q: How many samples will be needed to generate a valid model for use?

A: Based on our current test case, approximately 1000 samples that have been manually integrated, will be needed to generate a valid model for use.

Q: Will there be a difference in terms of the AI/ ML model due to differences between users’ methods? Or will all users share the same AI/ ML model?

A: Each AI/ML model deployed in MassHunter is unique to each customer according to their data analysis workflow for peak integration. The ML model is trained according to the methods each customer uses. Our system can support additional models specific to methods. Users within the same company can share to use the same trained model or they can choose to custom train multiple models definable by user, location (geographic), or lab.

To learn more about AI Peak Integration for MassHunter, contact your local Agilent representative or visit: https://www.agilent.com/mass-spec/ai-peak-integration