An Interview with AI About Its Potential Role in Separation Science

We interviewed an AI program (ChatGPT) for LCGC North America asking questions about AI and its role in various applications for separation science to include data analysis, and high performance liquid chromatography (HPLC), hydrophilic-interaction chromatography (HILIC), reversed-phase liquid chromatography (RPLC), liquid chromatography–mass spectrometry (LC–MS), gas chromatography–mass spectrometry (GC–MS), high resolution mass spectrometry (HRIM–MS), high resolution tandem mass spectrometry (HRMS/MS), and related topics. We asked 20 general questions for this interview, keeping in mind that our readers want to know all they can about how to use AI for their analytical chemistry interests. We also included a set of references and further reading for those wishing to look more deeply into the subject of AI and separation science.

It is important to note that a formal definition of AI does not always indicate the use of neural networks. Neural networks are one of the many techniques used in AI, but they are not the only ones. AI encompasses a wide range of techniques and approaches, including rule-based systems, decision trees, genetic algorithms, and others. While neural networks have gained prominence in recent years due to their ability to handle complex data and perform well on a variety of tasks, AI as a field is much broader and includes a range of techniques that do not rely on neural networks.

Robotic hand pressing a keyboard on a laptop 3D rendering | Image Credit: © sdecoret - stock.adobe.com

1. What is AI, and how does it differ from traditional computer programming?

Artificial Intelligence (AI) is a field of computer science that seeks to create intelligent machines capable of completing tasks that traditionally require human intelligence, including speech recognition, visual perception, natural language processing, and decision-making. Unlike traditional programming, which requires creating rules and instructions for a computer to solve a problem, AI systems learn from data and make decisions based on that data without being explicitly programmed. While traditional programming is deterministic and limited to solving specific tasks, AI systems can handle a wide range of problems, often involving unstructured data. They can learn from large datasets and find patterns that may be imperceptible to humans. Unlike traditional programming, AI can use techniques such as machine learning, deep learning, and neural networks to solve complex problems. Finally, AI systems can discover new insights, patterns, and knowledge without human intervention, which is different from traditional programming that requires a programmer to write code to solve a specific problem.

2. How is AI being used to improve data analysis in separation science?

AI is being used to improve data analysis in separation science in several ways (1–12). It can help automate the analysis of large datasets, improve the accuracy and consistency of data analysis, identify and prioritize areas for further study, optimize the design of separation experiments, better understand the complex relationships between different components in a mixture, and accelerate the development of new separation techniques and technologies. AI algorithms can be trained to recognize patterns in chromatography data, and can provide more accurate and consistent results than human analysts. By highlighting patterns or anomalies in the data, AI can identify new areas of research or potential problems. Additionally, AI can predict the most effective conditions for achieving a desired separation outcome, reducing the time and resources required for experimentation while improving overall data quality. Ultimately, AI is helping researchers in separation science gain new insights, discover new opportunities for exploration, and achieve faster and more reliable results.

3. Can you provide some examples of how AI has been used to optimize HPLC, HILIC, and RPLC methods?

AI has been used in various ways to optimize HPLC, HILIC, and RPLC methods (2,4,6–8). Machine learning algorithms can be used to optimize HPLC methods for separating mixtures of small molecules, peptides, and proteins. Deep learning can be used to identify the most informative features in complex chromatography data. AI algorithms can optimize chromatography methods based on multiple objectives, such as minimizing the analysis time while maximizing the resolution and peak capacity. Transfer learning can be used to optimize chromatography methods by leveraging knowledge learned from related separation techniques or datasets. AI algorithms can be used to detect anomalies and deviations from expected behavior in chromatography data, which can be indicative of problems with the instrument or the separation method. Overall, AI has the potential to significantly improve the efficiency, accuracy, and reliability of HPLC, HILIC, and RPLC methods, and to enable researchers to analyze complex mixtures more quickly and effectively.

4. How can AI be used to improve peak detection and quantification in chromatography?

AI can be used to improve peak detection and quantification in chromatography in several ways. First, machine learning algorithms can be used to train models that can accurately identify and differentiate between peaks in complex chromatography data, even in cases where there is significant overlap between peaks. Second, AI can be used to optimize the integration parameters used to quantify peaks, such as the baseline correction and peak width settings. Third, deep learning algorithms can be used to extract additional information from the chromatography data, such as peak shape and retention time, which can be used to improve peak quantification accuracy. Fourth, AI can be used to improve the accuracy of peak identification by comparing the retention times and spectral data of peaks against a database of known compounds. Fifth, AI can be used to identify and correct for baseline drift and other sources of noise in the chromatography data, which can improve the accuracy and consistency of peak detection and quantification. Finally, AI can be used to automatically detect and correct errors in the chromatography data, such as missing or mislabeled peaks, which can save time and improve the quality of the data.

5. What role does AI play in reducing method development time for chromatography techniques?

AI plays a significant role in reducing method development time for chromatography techniques. Firstly, it can be used to predict the most effective separation parameters for a given sample, reducing the need for extensive experimentation. Secondly, AI can be used to automate the process of method optimization, allowing researchers to rapidly test and refine multiple variables simultaneously. Thirdly, AI can be used to identify the most informative features in chromatography data, such as peak retention times and peak widths, which can be used to optimize separation conditions and minimize analysis time. Fourthly, AI can be used to simulate chromatography experiments, allowing researchers to test and compare different methods without the need for actual experiments. Fifthly, AI can be used to transfer knowledge learned from related separation techniques or datasets to optimize new methods. Finally, AI can be used to monitor and optimize the chromatography process in real-time, ensuring that the separation conditions remain optimal throughout the analysis. Overall, the use of AI can significantly reduce method development time, allowing researchers to develop and optimize new chromatography techniques quickly and efficiently.

6. How can AI be used to enhance selectivity and sensitivity in LC–MS and GC–MS methods?

AI can be used to enhance selectivity and sensitivity in LC–MS and GC–MS methods in several ways (9–11). Firstly, it can be used to optimize the selection of precursor and fragment ions for targeted MS/MS analysis, improving the selectivity and sensitivity of the method. Secondly, AI can be used to optimize the ionization conditions used in MS analysis, improving ionization efficiency and reducing interference from other compounds. Thirdly, AI can be used to predict the retention time of compounds in the chromatography column, allowing for the use of narrower chromatographic peaks and improved selectivity. Fourthly, AI can be used to automatically identify and remove background noise and interferences from MS data, improving sensitivity and reducing false positives. Fifthly, AI can be used to analyze large datasets of chromatography and MS data to identify correlations between compound properties and chromatographic behavior, which can be used to develop more selective and sensitive methods. Finally, AI can be used to monitor chromatographic performance in real-time, automatically adjusting separation conditions to maintain optimal sensitivity and selectivity throughout the analysis. Overall, the use of AI can significantly enhance the sensitivity and selectivity of LC–MS and GC–MS methods, allowing researchers to analyze complex mixtures more accurately and efficiently.

7. Can AI be used to predict retention times for compounds in chromatographic separations?

Yes, AI can be used to predict retention times for compounds in chromatographic separations (6,8). Machine learning algorithms can be trained using large datasets of chromatography data and chemical structure information to accurately predict the retention times of compounds based on their molecular features. These models can take into account a wide range of factors that influence retention, such as column chemistry, mobile phase composition, and temperature, allowing for highly accurate predictions. These retention time predictions can be used to optimize separation conditions, reduce analysis time, and improve peak identification. Furthermore, by analyzing the features that contribute to the retention time prediction, AI can be used to identify important molecular features that affect chromatographic behavior, providing valuable insights for method development and compound characterization.

8. How can AI be used to improve peak deconvolution and peak assignment in HRMS/MS and HRIM-MS?

AI can be used to improve peak deconvolution and peak assignment in HRMS/MS and HRIM-MS in several ways (12). Firstly, AI can be used to accurately deconvolve complex spectra into their component peaks, even in cases where the peaks are overlapping or have similar m/z values. Secondly, AI can be used to identify and assign the correct chemical formula to peaks based on their accurate mass, isotopic pattern, and fragmentation pattern information. Thirdly, AI can be used to predict the collision energy required for fragmentation of a given compound, improving the quality and specificity of MS/MS spectra. Fourthly, AI can be used to predict the retention time of compounds in chromatography, allowing for improved peak alignment and peak identification. Fifthly, AI can be used to automate the identification and quantification of metabolites in complex biological matrices, improving the speed and accuracy of metabolomics analyses. Finally, AI can be used to predict the fragmentation pattern of unknown compounds, enabling rapid and accurate compound identification even in cases where reference spectra are not available. Overall, the use of AI can significantly improve peak deconvolution and peak assignment in HRMS/MS and HRIM-MS, allowing for more accurate and efficient analyses of complex samples.

9. What are some of the challenges associated with using AI for separation science, and how can they be overcome?

There are several challenges associated with using AI for separation science. Firstly, the development of accurate and reliable AI models requires large amounts of high-quality data, which can be difficult to obtain in some cases. Secondly, the complexity of separation science data can make it difficult to identify relevant features and patterns for machine learning algorithms, requiring specialized expertise in both chemistry and data science. Thirdly, the interpretation of AI models can be challenging, as the underlying factors contributing to the model's predictions may not be immediately apparent. Fourthly, the use of AI may be limited by regulatory concerns and the need for validation and verification of the models. To overcome these challenges, efforts should be made to increase the availability of high-quality data, develop specialized tools for data preprocessing and feature extraction, and enhance collaboration between chemists and data scientists. Additionally, efforts should be made to improve the transparency and interpretability of AI models, and to establish clear guidelines for the validation and verification of AI models in separation science.

10. How can AI be used to improve the reliability and reproducibility of chromatographic methods?

AI can be used to improve the reliability and reproducibility of chromatographic methods in several ways. Firstly, AI models can be used to predict the optimal chromatographic conditions for a given separation, reducing the need for trial-and-error optimization and improving the consistency of results. Secondly, AI can be used to automate the identification and correction of systematic errors in chromatographic data, such as baseline drift or column degradation, improving the accuracy and reproducibility of results. Thirdly, AI can be used to monitor chromatographic performance in real-time, allowing for rapid detection of any deviations from the expected behavior and enabling timely corrective actions. Fourthly, AI can be used to optimize sample preparation protocols, reducing variability in sample matrix effects and improving the accuracy and precision of quantitative analyses. Fifthly, AI can be used to analyze and compare large datasets of chromatographic data, enabling the identification of trends and patterns that may not be immediately apparent to human analysts. Finally, AI can be used to develop predictive models for chromatographic performance, allowing for more accurate predictions of separation behavior under a wide range of conditions. Overall, the use of AI has the potential to significantly improve the reliability and reproducibility of chromatographic methods, leading to more accurate and consistent results.

11. Can AI be used to improve the accuracy and precision of mass spectrometry-based methods?

Yes, AI can be used to improve the accuracy and precision of mass spectrometry-based methods by automating the process of identifying and quantifying compounds in complex matrices. By leveraging machine learning algorithms and other AI techniques, researchers can quickly and accurately identify and quantify compounds in complex mixtures, improving the accuracy and precision of the method.

12. What are some of the ethical considerations associated with using AI in separation science?

Some of the ethical considerations associated with using AI in separation science include data privacy, bias in algorithmic decision-making, and the impact of automation on employment in the field. Researchers using AI in separation science should be mindful of these considerations and take steps to ensure that their use of AI is transparent, unbiased, and respectful of human rights and privacy.

13. Can AI be used to identify unknown compounds in complex matrices?

Yes, AI can be used to identify unknown compounds in complex matrices by leveraging machine learning algorithms and other AI techniques. By training models on large data sets of known compounds and their properties, researchers can develop models that can accurately predict the identity of unknown compounds based on their mass spectra, chromatographic retention times, and other properties.

14. How can AI be used to improve the speed and efficiency of method development for separation science?

AI can be used to improve the speed and efficiency of method development for separation science by automating the process of generating and testing chromatographic conditions. By leveraging machine learning algorithms and other AI techniques, researchers can quickly and efficiently generate and test thousands of different chromatographic conditions to identify the optimal parameters for a given separation.

15. Can AI be used to identify impurities or contaminants in samples?

Yes, AI can be used to identify impurities or contaminants in samples by leveraging machine learning algorithms and other AI techniques. By training models on large data sets of known contaminants and their properties, researchers can develop models that can accurately predict the presence of impurities or contaminants in a sample based on their mass spectra, chromatographic retention times, and other properties.

16. How can AI be used to improve the reliability and accuracy of peak integration in chromatography?

AI can be used to improve the reliability and accuracy of peak integration in chromatography by automating the process of identifying and quantifying peaks in chromatograms. By leveraging machine learning algorithms and other AI techniques, researchers can quickly and accurately identify and quantify peaks in chromatograms, reducing the likelihood of errors or variability due to human error or inconsistency.

17. Can AI be used to optimize sample preparation methods for chromatography techniques?

Yes, AI can be used to optimize sample preparation methods for chromatography techniques by automating the process of selecting optimal extraction or cleanup methods. By leveraging machine learning algorithms and other AI techniques, researchers can quickly generate and test different sample preparation methods to identify the optimal parameters for a given separation.

18. How can AI be used to improve the reproducibility of chromatographic methods across different laboratories?

AI can be used to improve the reproducibility of chromatographic methods across different laboratories by developing standardized protocols and using machine learning algorithms to optimize and validate these methods. By developing standardized protocols and using machine learning algorithms to identify the optimal chromatographic conditions for a given separation, researchers can ensure that their methods are reproducible and reliable across different laboratories.

19. How can AI be used to identify and classify different classes of compounds in complex matrices?

AI can be used to identify and classify different classes of compounds in complex matrices by leveraging machine learning algorithms and other AI techniques to train models on large data sets of known compounds and their properties. By identifying patterns in the data, researchers can develop models that can accurately predict the identity and class of unknown compounds based on their mass spectra, chromatographic retention times, and other properties.

20. What are some of the limitations and challenges associated with using AI in separation science, and how can they be overcome?

Some of the limitations and challenges associated with using AI in separation science include the need for large and high-quality data sets, the complexity of the algorithms required, and the need for expertise in both analytical chemistry and computer science. These challenges can be overcome by developing robust and comprehensive data sets, collaborating with experts in both fields, and leveraging open-source software and tools to simplify the implementation of AI in analytical chemistry. Additionally, it is important to be mindful of ethical considerations and to ensure that AI is used in a transparent and unbiased manner.

References

(1) Grinias, J. P. Navigating the Future of Separation Science Education: A Perspective. Chromatographia 2022, 85 (8), 681–688. DOI: https://doi.org/10.1007/s10337-022-04182-8

(2) Usman, A. G.; Işik, S.; Abba, S. I. Qualitative Prediction of Thymoquinone in the High‐Performance Liquid Chromatography Optimization Method Development Using Artificial Intelligence Models Coupled With Ensemble Machine Learning. Sep. Sci. Plus 2022, 5 (10), 579–587. DOI: https://doi.org/10.1002/sscp.202200071

(3) Houhou, R.; Bocklitz, T. Trends in Artificial Intelligence, Machine Learning, and Chemometrics Applied to Chemical Data. Anal. Sci. Adv. 2021, 2 (3-4), 128–141. DOI: https://doi.org/10.1002/ansa.202000162

(4) Abba, S. I.; Usman, A. G.; and Selin, I. Simulation for Response Surface in the HPLC Optimization Method Development Using Artificial Intelligence Models: A Data-Driven Approach. Chemom. Intell. Lab. Syst. 2020, 201, 104007. DOI: https://doi.org/10.1016/j.chemolab.2020.104007

(5) Mann, M.; Kumar, C.; Zeng, W. F.; Strauss, M. T. Artificial Intelligence for Proteomics and Biomarker Discovery. Cell Syst. 2021, 12 (8), 759–770. DOI: https://doi.org/10.1016/j.cels.2021.06.006

(6) Gritti, F. Perspective on the Future Approaches to Predict Retention in Liquid Chromatography. Anal. Chem. 2021, 93 (14), 5653–5664. DOI: https://doi.org/10.1021/acs.analchem.0c05078

(7) Anderson, H. E.; Liden, T.; Berger, B. K.; Zanella, D.; Manh, L, H.; Wang, S.; Schug, K. A. Profiling of Contemporary Beer Styles Using Liquid Chromatography Quadrupole Time-of-Flight Mass Spectrometry, Multivariate Analysis, and Machine Learning Techniques. Anal. Chim. Acta 2021, 1172, 338668. DOI: https://doi.org/10.1016/j.aca.2021.338668

(8) Win, Z. M.; Cheong, A. M.; Hopkins, W. S. Using Machine Learning To Predict Partition Coefficient (Log P) and Distribution Coefficient (Log D) with Molecular Descriptors and Liquid Chromatography Retention Time. J. Chem. Inf. Model. 2023, 63 (7), 1906–1913. DOI: https://doi.org/10.1021/acs.jcim.2c01373

(9) Zhang, L.;, Ma, F.; Qi, A.; Liu, L.; Zhang, J.; Xu, S.; Zhong, Q.; Chen, Y.; Zhang, C. Y.; Cai, C. Integration of Ultra-High-Pressure Liquid Chromatography–Tandem Mass Spectrometry With Machine Learning for Identifying Fatty Acid Metabolite Biomarkers of Ischemic Stroke. ChemComm. 2020, 56 (49), 6656–6659. DOI: https://doi.org/10.1039/D0CC02329A

(10) Zeng X.; Ma, B. MSTracer: A Machine Learning Software Tool for Peptide Feature Detection from Liquid Chromatography–Mass Spectrometry Data. J. Proteome Res. 2021, 20 (7), 3455–3462. DOI: https://doi.org/10.1021/acs.jproteome.0c01029

(11) Brendel, R.; Schwolow, S.; Rohn, S.; Weller, P. Volatilomic Profiling of Citrus Juices by Dual-Detection HS-GC-MS-IMS and Machine Learning—An Alternative Authentication Approach. J. Agric. Food Chem. 2021, 69 (5), 1727–1738. DOI: https://doi.org/10.1021/acs.jafc.0c07447

(12) Kantz, E. D.; Tiwari, S.; Watrous, J. D.; Cheng, S.; Jain, M. Deep Neural Networks for Classification of LC-MS Spectral Peaks. Anal. Chem. 2019,91 (19), 12407–12413. DOI: https://doi.org/10.1021/acs.analchem.9b02983

About the Author

Jerome Workman, Jr. is the Senior Technical Editor of LCGC and Spectroscopy. Direct correspondence to: jworkman@mjhlifesciences.com