OR WAIT null SECS
First defined by the Centers for Disease Control and Prevention (CDC) following the 1990–1991 Gulf War, established medical diagnoses, laboratory tests, and hypothesis-driven research have failed to explain the multifacted symptomatology of Gulf War illness (GWI), a chronic illness with no known validated biomarkers. Noureddine Melikechi of the Department of Physics and Applied Physics at the University of Massachusetts (Lowell, MA) saw an urgent need for the development of an untargeted and unbiased method to distinguish GWI patients from non-GWI patients; he and his associates utilized laser-induced breakdown spectroscopy (LIBS) in their efforts to meet that need. Melikechi spoke with Spectroscopy about this work, and his research paper that has resulted.
In one of your recent papers (1), you focus on the diagnosis of Gulf War illness (GWI). In that paper you used laser-induced breakdown spectroscopy (LIBS) to measure blood samples in order to differentiate between GWI patients and non-GWI patients. Would you briefly describe the results of this research and what you have learned?
Gulf War Illnessaffects a quarter of the nearly 700,000 veterans who served in the 1990–1991 Gulf War. It is a chronic illness with multiple symptoms that include both physical and emotional disorders. Using a LIBS-liquid biopsyapproach similar to that of our previous studies, we searched for spectral signatures that can differentiate blood plasma of cancerous and non-cancerous mice. We have also shown that this technique can be used to differentiate blood plasma of healthy patients from patients known to suffer from Alzheimer’s disease. For our GWI work, we examined the LIBS signatures of blood plasma acquired from GWI patients and non-GWI patients. We used various machine learning techniques for the analysis of the spectra obtained. By comparing the LIBS spectra obtained from blood plasma of GWI patients and non GWI patients, we generated a classification model which was used to conduct a blind test. This test yielded a classification accuracy of the unknown cases of 90% with 100% sensitivity, and 83% specificity. These are encouraging results, but it is important to note that this initial study was performed with a smaller number of samples than we would have liked. We plan to examine samples from a larger number of patients in the next phase of our work.
You also have used LIBS to research the diagnosis of melanoma in biomedical fluids (2), Alzheimer's disease (3), and epithelial ovarian cancer (EOC) (4). Which application looks the most promising, and why?
My research group has been working on the development of minimally invasive, multidisciplinary techniques for the detection, screening, and diagnosis of cancer with special emphasis on EOC and melanoma, as well as on early signs of Alzheimer’s disease using (principally) blood plasmas. For these diseases, we have obtained discrimination accuracies between diseased and healthy controls that approach, and in some cases surpass, 80%. Other researchers have also been able to apply similar approaches and obtain encouraging results, for example, on the detection of multiple myeloma in humans (https://doi.org/10.1364/BOE.421333). It is too early to know which, if any, application is more promising as much more work needs to be done. Can LIBS alone or integrated with other analytical techniques contribute to the screening and possibly diagnosis of cancers and neurological diseases? Much work needs to be done before this question can be answered with more certainty.
What benefits have you seen using LIBS spectroscopy over other spectroscopic techniques in diagnosing various types of illness?
There are several spectroscopic techniques that can be envisaged, such as Raman spectroscopy, Fourier transform infrared spectroscopy–attenuated total reflectance (FTIR-ATR), and inductively coupled plasma mass spectrometry(ICP-MS). We have selected to use LIBS because it is a multi-elemental analysis technique that can be performed with very limited sample preparation and can provide qualitative and quantitative information for a great variety of samples, including biomedical fluids. Few other spectroscopic techniques present such advantages. However, LIBS also has limitations, for example in sensitivity. It may be that integrating atomic and ionic information obtained using LIBS with molecular information obtained using Raman or FTIR-ATR or preparing samples prior to making measurements, such as by using Tag-LIBS, will provide better results.
What are the biggest challenges that you have encountered in developing LIBS methods for diagnostic applications? What options or alternative developments are available to overcome these challenges or to improve your approach?
The main challenges are not specific to developing LIBS methods. First, this work requires the commitment and engagement of an interdisciplinary team dedicated to examining, what may often be an old problem, with new perspectives. Only then, can the interdisciplinary team of researchers synergistically take advantage of the knowledge and experience of each. This is a challenge as we are, typically, not trained to work in interdisciplinary teams. Second, it is critical to have access to biomedical samples of “good quality” and in sufficient quantity to be able to acquire “enough” spectroscopic data. As LIBS is essentially sensitive to practically all chemical elements, significant attention must be given to ensure that all biomedical samples are treated, during and post collection, the same way. Failure to do this can result in the introduction of external chemical elements in one or more of the samples that can influence the measurements and hence the final analysis and conclusions. Third, it is important to keep an open mind and learn from what the data tells us and not be blinded by pre-conceived and possibly hidden assumptions.
The abundance of biomedical data, such as spectroscopic and imaging data, and the development of powerful new computational resources may allow for early detection of diseases that were previously impossible to attain. A key step is to clearly and unambiguously identify the atomic and molecular features that are behind the differences observed in the spectra and to understand at the fundamental level the reasons for the existence of such differential spectroscopic signatures. We are in the process of examining our data to seek such information.
What sort of feedback did you receive from these papers and your reported results?
The fact that, as interdisciplinary teams, we are looking at cancers, Alzheimer’s disease, and gulf war illness with researchers from different institutions is a sign that there is an interest in this approach.
Are there any additional analytical approaches or methods that you are planning to explore? Would you describe your use of specialized chemometrics and machine learning and how these techniques are an advantage in data analysis?
We are refining our analytical approaches by investigating the use of powerful, novel, as well as established, machine learning tools. Many of these tools have not been investigated fully yet in this area of research. We are also looking into extracting spectral features from the LIBS data that contribute significantly to the differentiation and classification of the blood plasma samples. Such features will be used to train our models and to identify the associated chemical elements.
What are your next steps regarding your LIBS research?
The studies we have conducted thus far are focused on distinguishing the LIBS spectra of the blood plasma of healthy patients with patients identified as having only one disease. Reality is more complex. A richer, more comprehensive clinical study that involves a larger number of patients with multiple conditions is needed. Such a study will help evaluate the specificity and sensitivity accuracies of the approach in scenarios that are closer to clinical settings. It may be useful to consider combining atomic and ionic information with molecular information of blood plasma samples and possibly proteomics. Together these tools may provide clues that help identify new biomarkers for cancers or neurological diseases. As useful as the analysis of spectral data with machine learning techniques is at distinguishing diseased and healthy biomedical samples, much work is needed to better understand the onset of diseases at the atomic and molecular levels prior to the appearance of clinical symptoms.
(1) R. Gaudiuso, S. Chen, E. Kokkotou, L. Conboy, E. Jacobson, E.B. Hanlon, and N. Melikechi, Appl. Spectrosc. (2021) https://doi.org/10.1177/00037028211042049
(2) R. Gaudiuso, E. Ewusi-Annan, N. Melikechi, X. Sun, B. Liu, L.F. Campesato, and T. Merghoub, Spectrochim. Acta, Part B 146, 106-114 (2018).
(3) R. Gaudiuso, E. Ewusi-Annan, W. Xia, and N. Melikechi, Spectrochim. Acta, Part B 171, 105931 (2020).
(4) N. Melikechi, Y. Markushin, D.C. Connolly, J. Lasue, E. Ewusi-Annan, and S. Makrogiannis, Spectrochim. Acta, Part B 123, 33-41 (2016)
(5) Y. Markushin, P. Sivakumar, D.C. Connolly, and N. Melikechi, Analytical and Bioanalytical Chemistry, 407, 1849-1855 (2015)
Noureddine Melikechi is a Professor of Physics and serves as the Dean of the Kennedy College of Sciences at the University of Massachusetts Lowell. In recent years, Noureddine’s research group has been working on the development of approaches for the early detection of laser spectroscopic signatures of cancers and Alzheimer’s disease. In addition, as a member of NASA’s Mars missions, Noureddine contributes to the analysis of LIBS data collected by the ChemCam and the SuperCam instruments on board the Curiosity and Perseverance rovers. A native of Algeria, Noureddine received his Diplôme d'Études Supérieures in Physics from the University of Sciences and Technology Houari Boumediene, Algeria, and his M.Sc. and D.Phil., both in Physics, from the University of Sussex. Noureddine is a Fellow of Optica, the America Physical Society and the American Association for the Advancement of Science.