A model of clinical and environmental risk based on artificial intelligence

Cover Page

Cite item

Full Text

Open Access Open Access
Restricted Access Access granted
Restricted Access Subscription or Fee Access

Abstract

The exposome concept emphasizes the importance of taking into account the cumulative effects on the body during life, including environmental pollutants. The influence of these factors may increase the risk of developing chronic diseases.

Objective. Development of a clinical and environmental risk assessment model that allows predicting the level of sensitivity to chemicals based on measured biomarkers and the assessment of the Quick Environmental Exposure and Sensitivity Inventory (QEESI) index.

Material and methods. The study included 71 patients (32 men and 39 women) aged 18–65 years. The level of chemical sensitivity exceeded 20 points on the QEESI scale. The data was analyzed using Python and statistical libraries. A machine learning model (Random Forest Regressor) was created to assess clinical and environmental risk.

Results. Significant deviations of biomarkers (alanine aminotransferase, aspartate aminotransferase, low-density lipoproteins) from the normal distribution required the use of mathematical transformations. The model showed good predictive abilities, despite the high values of MAE and RMSE. The cubic and quadratic shape of bilirubin and the scale of symptoms were the most significant factors.

Conclusions. The developed model based on the Random Forest Regressor algorithm has shown high accuracy in predicting clinical and environmental risk. Optimization of hyperparameters, preprocessing of data using mathematical transformations (logarithmic, square, cubic), the use of the feature_importances_ attribute allowed a deeper understanding of the impact of environmental factors on health.

Full Text

Restricted Access

About the authors

V. V. Onopriev

Kuban State Medical University

Email: bolotowa_e@mail.ru
ORCID iD: 0000-0002-6676-5247
SPIN-code: 5631-6667

MD

Russian Federation, Krasnodar

E. V. Bolotova

Kuban State Medical University

Email: bolotowa_e@mail.ru
ORCID iD: 0000-0001-6257-354X
SPIN-code: 4412-3236

MD, Professor

Russian Federation, Krasnodar

A. V. Dudnikova

Kuban State Medical University

Email: bolotowa_e@mail.ru
ORCID iD: 0000-0003-2601-7831
SPIN-code: 7480-1992

Candidate of Medical Sciences

Russian Federation, Krasnodar

L. V. Batrakova

Kuban State Medical University

Author for correspondence.
Email: bolotowa_e@mail.ru
ORCID iD: 0000-0002-3688-6064
Russian Federation, Krasnodar

A. G. Abramenko

Kuban State Medical University

Email: bolotowa_e@mail.ru
ORCID iD: 0009-0007-6649-8576
SPIN-code: 5540-2472
Russian Federation, Krasnodar

References

  1. Bray R., Wang Y., Argiropoulos N. et al. The Impact of COVID-19 Health Measures on Adults With Multiple Chemical Sensitivity: Cross-Sectional Study. JMIR Form Res. 2024; 8: e48434. doi: 10.2196/48434
  2. Del Casale A., Ferracuti S., Mosca A. et al. Multiple Chemical Sensitivity Syndrome: A Principal Component Analysis of Symptoms. Int J Environ Res Public Health. 2020; 17 (18): 6551. doi: 10.3390/ijerph17186551
  3. Sanchez C.A., Lozada-Urbano M., Ospina E.E. Environmental exposure to multiple chemical elements in Peruvian populations: a review of selected studies. F1000Research. 2024; 13: 240. doi: 10.12688/f1000research.143130.1
  4. Utsav A., Mani H., Kumar M. еt al. Exposome analysis in toxicology: A comprehensive review. Journal of Toxicological Studies. 2023; 1 (1): 267. doi: 10.59400/jts.v1i1.267
  5. Cave M., Appana S., Patel M. еt al. Polychlorinated biphenyls, lead, and mercury are associated with liver disease in American adults: NHANES 2003-2004. Environ Health Perspect. 2010; 118 (12): 1735–42. doi: 10.1289/ehp.1002720
  6. Younossi Z.M., Stepanova M., Afendy M. еt al. Changes in the prevalence of the most common causes of chronic liver diseases in the United States from 1988 to 2008. Clin Gastroentero. Hepatol. 2011; 9 (6): 524–530.e1. doi: 10.1016/j.cgh.2011.03.020
  7. Lee H., Kim Y., Sim C.S., еt al. Associations between blood mercury levels and subclinical changes in liver enzymes among South Korean general adults: analysis of 2008-2012 Korean National Health and Nutrition Examination Survey data. Environ Res. 2014; 130: 14–9. doi: 10.1016/j.envres.2014.01.005
  8. Min Y.S., Lim H.S., Kim H. Biomarkers for polycyclic aromatic hydrocarbons and serum liver enzymes. Am J Ind Med. 2015; 58 (7): 764–72. doi: 10.1002/ajim.22463
  9. Lee D.H., Ha M.H., Kim J.H. еt al. Gamma-glutamyltransferase and diabetes—a 4 year follow-up study. Diabetologia. 2003; 46 (3): 359–64. doi: 10.1007/s00125-003-1036-5
  10. Pompella A., Emdin M., Franzini M. еt al. Serum gamma-glutamyltransferase: linking together environmental pollution, redox equilibria and progression of atherosclerosis? Clin Chem Lab Med. 2009; 47 (12): 1583–4. doi: 10.1515/CCLM.2009.350
  11. Go Y.M., Walker D.I., Liang Y. еt al. Reference Standardization for Mass Spectrometry and High-resolution Metabolomics Applications to Exposome Research. Toxicol Sci. 2015; 148 (2): 531–43. doi: 10.1093/toxsci/kfv198
  12. Vrijheid M., Fossati S., Maitre L. еt al. Early-Life Environmental Exposures and Childhood Obesity: An Exposome-Wide Approach. Environ Health Perspect. 2020; 128 (6): 67009. doi: 10.1289/EHP5975
  13. Matus P., Sepúlveda-Peñaloza A., Page K. еt al. The Chilean exposome-based system for ecosystems (CHiESS): a framework for national data integration and analytics platform. Front Public Health. 2024; 12: 1407514. doi: 10.3389/fpubh.2024.1407514
  14. Wang K., Liu P., Sun F. et al. Progress in Realizing the Value of Ecological Products in China and Its Practice in Shandong Province. Sustainability. 2023; 15: 9480. doi: 10.3390/su15129480
  15. Shahbazi Z., Camacho M., Ruiz E. et al. Machine learning-based prediction of diabetes risk by combining exposome and electrocardiographic predictors. In: 18th International Symposium on Medical Information Processing and Analysis. 2023 Mar; Vol. 12567: pp. 277–85.
  16. Higueras-Serrano I., Cerván-Martin M., González Muñoz S. et al. Influence of a dynamic exposome on the genetic component of male infertility. Hum Reprod. 2024; 39 (3): 612–22. doi: 10.1093/humrep/deae007
  17. Ashford N.A., Miller C.S. 1998: Appendix C: Environmental Exposure and Sensitivity Inventory: Chemical Application of QEESI for Japanese population. New York: John Wiley and Sons, pp. 371–84.
  18. Болотова Е.В., Дудникова А.В., Оноприев В.В. и др. Валидация русскоязычной версии опросника QEESI для оценки синдрома множественной химической чувствительности. Врач. 2024; 35 (7): 57–64 [Bolotova E., Dudnikova A., Onopriev V. et al. Validation of the russian-language version of the QEESI questionnaire for the assessment of multiple chemical sensitivity syndrome. Vrach. 2024; 35 (7): 57–64 (in Russ.)]. doi: 10.29296/25877305-2024-07-10
  19. Miller C.S., Prihoda T.J. A controlled comparison of symptoms and chemical intolerances reported by Gulf War veterans, implant recipients and persons with multiple chemical sensitivity. Toxicol Ind Health. 1999; 15 (3-4): 386–97. doi: 10.1177/074823379901500312
  20. Huang H., Chen H.M., Lin W.W. et al. Employing feature engineering strategies to improve the performance of machine learning algorithms on echocardiogram dataset. Digit Health. 2023; 9: 20552076231207589. doi: 10.1177/2055207623120758
  21. Feurer M., Hutter F. Hyperparameter Optimization. In: Hutter F., Kotthoff L., Vanschoren J. (eds). Automated Machine Learning. The Springer Series on Challenges in Machine Learning. Springer, Cham, 2019. doi: 10.1007/978-3-030-05318-5_1
  22. Brownlee J. Machine learning mastery with Python: understand your data, create accurate models and work projects end-to-end. Edition: v1.20, 2021; 169 р.

Supplementary files

Supplementary Files
Action
1. JATS XML
Download
2. Fig. 1. Comparison of actual and predicted values: a – train data; б – test data

Download (81KB)
Indexing metadata
3. Fig. 2. Significance of each attribute for clinical and environmental risk assessment

Download (335KB)
Indexing metadata
4. Fig. 3. ROC curves for training and test samples: light gray line - training sample (AUC1 = 0.94), dark gray line - test sample (AUC2 = 0.93), dashed line shows the random selection line

Download (85KB)
Indexing metadata

Copyright (c) 2025 Russkiy Vrach Publishing House