DAHA AZ DENEME GERÇEKLEŞTİRMEK MÜMKÜN MÜ: BAYESIAN YAKLAŞIMLA YÜKSEK BASINÇ İŞLEMLERİ İÇİN BAKTERİYEL HAYATTA KALMA/ÖLÜM OLASILIĞININ TAHMİNİ?

Abstract

Mevcut çalışmada, farklı koşullar altında yüksek hidrostatik basınç işlemine tabi tutulan bakterilerin hayatta kalma/ölüm olasılığını tahmin etmek için Bayesian Logistic Regression'a (BLR) dayalı bir model geliştirilmiştir. Bu amaçla Listeria monocytogenes (fosfatla tamponlanmış tuzlu su çözeltisi içinde) ve Cronobacter sakazakii (triptik soya broth ve bebek maması formülasyonu) bakterileri için daha önce yayımlanmış verilerden faydalanılmış olup, proses değişkenleri basınç, sıcaklık, ortamın pH değeri, ilk aşılama ve işlem süresidir. BLR kullanım olasılığının yanı sıra, veri bölme oranları değiştirilerek örneklem büyüklüğünün ve verilerdeki vaka sıklığının etkileri değerlendirilmiştir. Sonuç olarak BLR model tahminlerinin hem deneysel verilerle hem de frekansçı lojistik regresyon modelleriyle tutarlı olduğu gözlenmiştir. Örneklem boyutu küçüldükçe bazı aşırı uyum sorunları ortaya çıksa da, BLR, frekansçı yaklaşımının gerektirdiğinden daha az sayıda deneysel veriye ile (yaklaşık 50 deneysel örnek) güvenilir olasılık modelleri üretebilmektedir. Dahası BLR, nokta tahminleri yerine parametreler ve kestirimler için sonsal dağılımlar sunmaktadır. Bu nedenle mevcut çalışmada, BLR'nin frekansçı yaklaşıma göre daha az deneysel veri gereksinimiyle mikroorganizmaların uygulanan yüksek basınç işlemlerinden sonra hayatta kalma/ölme olasılık kestirimleri için yararlı bir araç olabileceği, eksik gözlemleri ve dengesiz veri setlerini yönetme kabiliyetine sahip olduğu gösterilmiştir. Bu sonuçların ışığında, BLR yaklaşımına uygun yeni deneme tasarımları ile, deneysel çalışmalarda zamandan ve maliyetten tasarruf sağlanması ve gıda endüstrisi için daha ayrıntılı güvenlik riski değerlendirmesi mümkün olabilir.

Keywords

• Yüksek hidrostatik basınç
• Öngörücü mikrobiyoloji
• Lojistik regresyon
• Listeria monocytogenes
• Cronobacter sakazakii

IS IT POSSIBLE TO MAKE FEWER EXPERIMENTS: PREDICTION OF BACTERIAL SURVIVAL/DEATH PROBABILITY FOR HIGH-PRESSURE PROCESSING WITH THE BAYESIAN APPROACH?

Abstract

In the present study, a model based on Bayesian Logistic Regression (BLR) was developed to predict the probability of bacterial survival/death treated with high-hydrostatic pressure under different conditions. Previously published data for Listeria monocytogenes in phosphate-buffered saline and Cronobacter sakazakii in trypticase soy broth and infant formula were used where the process variables were pressure, temperature, medium pH, initial inoculum and processing time. Along with the using possibility of BLR, effects of introduced sampling size by changing data split ratio and case prevalence were assessed. The BLR model predictions were consistent with both experimental data and the frequentist logistic regression models. Although some overfitting problems arise as the sampling size decrease, BLR can produce reliable probability models with a smaller number of experimental data (about 50 experimental samples) than the frequentist approach requires. Moreover, instead of a point estimate, BLR offers a posterior distribution for parameters and predictions. So the present study has indicated that BLR can be a useful tool to describe the survival/death of microorganisms after high-pressure processes with less experimental data requirement than the frequentist approach and also with the ability to handle missing observation and imbalanced dataset. In the light of these outcomes, the design of new experiments according to BLR, save on time and costs for experimental studies and more detailed safety risk assessment may be feasible for the food industry.

Keywords

• High hydrostatic pressure
• Predictive microbiology
• Logistic regression
• Listeria monocytogenes
• Cronobacter sakazakii

References

1. Baratloo, A., Hosseini, M., Negida, A., El Ashal, G., (2015). Part 1: Simple definition and calculation of accuracy, sensitivity and specificity. Emergency 3(2), 48-49.
2. Bilbao-Sáinz, C., Younce, F.L., Rasco, B., Clark, S., (2009). Protease stability in bovine milk under combined thermal-high hydrostatic pressure treatment. Innovative Food Science & Emerging Technologies 10(3), 314-320.
3. Bland, J.M., Altman, D.G., (2000). The odds ratio. BMJ 320(7247), 1468.
4. Butz, P., Fernández Garcı́a, A., Lindauer, R., Dieterich, S., Bognár, A., Tauscher, B., (2003). Influence of ultra high pressure processing on fruit and vegetable products. Journal of Food Engineering 56(2), 233-236.
5. Buzrul, S., (2014). Multi-pulsed high hydrostatic pressure inactivation of microorganisms: A review. Innovative Food Science & Emerging Technologies 26, 1-11.
6. Buzrul, S., (2017). Evaluation of different dose-response models for high hydrostatic pressure inactivation of microorganisms. Foods 6(9), 79.
7. Buzrul, S., (2019). High hydrostatic pressure inactivation of microorganisms: A probabilistic model for target log-reductions. International Journal of Food Microbiology 309, 108330.
8. Buzrul, S., Alpas, H., (2004). Modeling the synergistic effect of high pressure and heat on inactivation kinetics of Listeria innocua: a preliminary study. FEMS Microbiology Letters 238(1), 29-36.

9. Buzrul, S., Alpas, H., Bozoglu, F., (2005). Use of Weibull frequency distribution model to describe the inactivation of Alicyclobacillus acidoterrestris by high pressure at different temperatures. Food Research International 38(2), 151-157.
10. Buzrul, S., Alpas, H., Largeteau, A., Demazeau, G., (2008). Modeling high pressure inactivation of Escherichia coli and Listeria innocua in whole milk. European Food Research and Technology 227(2), 443-448.
11. Campus, M., (2010). High pressure processing of meat, meat products and seafood. Food Engineering Reviews 2(4), 256-273.
12. Carroll, C., (2019). Pragmatic probabilistic programming: Parameter adaptation in PyMC3.
13. du Prel, J.-B., Hommel, G., Röhrig, B., Blettner, M., (2009). Confidence interval or p-value?: part 4 of a series on evaluation of scientific publications. Deutsches Arzteblatt international 106(19), 335-339.
14. Ferrari, G., Maresca, P., Ciccarone, R., (2010). The application of high hydrostatic pressure for the stabilization of functional foods: Pomegranate juice. Journal of Food Engineering 100(2), 245-253.
15. Gudbjornsdottir, B., Jonsson, A., Hafsteinsson, H., Heinz, V., (2010). Effect of high-pressure processing on Listeria spp. and on the textural and microstructural properties of cold smoked salmon. LWT - Food Science and Technology 43(2), 366-374.
16. Hajmeer, M., Basheer, I., (2002). A probabilistic neural network approach for modeling and classification of bacterial growth/no-growth data. Journal of Microbiological Methods 51(2), 217-226.
17. Hite, B.H., (1899). The effects of pressure in the preservation of milk. West Virginia Agricultural and Forestry Experiment Station Bulletins 58, 15-35.
18. Hokmollahi, F., Ehsani, M., (2017). High pressure processing and its application in cheese manufacturing: a review. Journal of Food Biosciences and Technology 7, 57-66.
19. James, G., Witten, D., Hastie, T., Tibshirani, R., (2013). An Introduction to statistical learning with applications in R. Springer-Verlag New York.
20. Khan, M.A., Ali, S., Abid, M., Ahmad, H., Zhang, L., Tume, R.K., Zhou, G., (2014). Enhanced texture, yield and safety of a ready-to-eat salted duck meat product using a high pressure-heat process. Innovative Food Science & Emerging Technologies 21, 50-57.
21. Koehrsen, W., (2018). Bayesian linear regression in python: using machine learning to predict student grades part 2. Medium.
22. Korner-Nievergelt, F., Roth, T., Von Felten, S., Guélat, J., Almasi, B., Korner-Nievergelt, P., (2015). Bayesian data analysis in ecology using linear models with R, BUGS, and Stan. Academic Press.
23. Koseki, S., Matsubara, M., Yamamoto, K., (2009). Prediction of a required log reduction with probability for Enterobacter sakazakii during high-pressure processing, using a survival/death interface model. Applied and Environmental Microbiology 75(7), 1885.
24. Koseki, S., Yamamoto, K., (2007). Modelling the bacterial survival/death interface induced by high pressure processing. International Journal of Food Microbiology 116(1), 136-143.
25. Lu, S., (2019). Do you know credible interval? Medium.
26. McElreath, R., (2016). Statistical rethinking. A bayesian course with examples in R and stan. CRC Press, Boca Raton.
27. O'Brien, S.M., Dunson, D.B., (2004). Bayesian Multivariate Logistic Regression. Biometrics 60(3), 739-746.
28. Pacifico, A., (2021). Robust open Bayesian analysis: Overfitting, model uncertainty, and endogeneity issues in multiple regression models. Econometric Reviews 40(2), 148-176.
29. PyMC3, (2018). GLM: Linear regression. The PyMC Development Team.
30. Rahman, H.A., Wah, Y.B., Huat, O.S., (2020). Predictive performance of logistic regression for imbalanced data with categorical covariate. pertanika journal of science and technology 28.
31. Ratkowsky, D.A., (2004). Model fitting and uncertainty, in: McKellar, R.C., Lu, X. (Ed.), Modeling microbial responses in food. CRC Press, Boca Raton, FL.
32. Salvatier, J., Wiecki, T.V., Fonnesbeck, C., (2016). Probabilistic programming in Python using PyMC3. PeerJ Computer Science 2, e55.
33. Serment-Moreno, V., Barbosa-Cánovas, G., Torres, J.A., Welti-Chanes, J., (2014). High-pressure processing: kinetic models for microbial and enzyme inactivation. Food Engineering Reviews 6(3), 56-88.
34. Stoica, M., Mihalcea, L., Borda, D., Alexe, P., (2013). Non-thermal novel food processing technologies. An overview. Journal of Agroalimentary Processes and Technologies 19(2), 212-217.
35. Szumilas, M., (2010). Explaining odds ratios. Journal of the Canadian Academy of Child and Adolescent Psychiatry 19(3), 227-229.
36. UCLA, (2016). How do i interpret odds ratios in logistic regression? UCLA: Statistical Consulting Group.
37. van Boekel, M.A.J.S., (2020). On the pros and cons of Bayesian kinetic modeling in food science. Trends in Food Science & Technology 99, 181-193.
38. Vehtari, A., Gelman, A., Gabry, J., (2017). Practical Bayesian model evaluation using leave-one-out cross-validation and WAIC. Statistics and Computing 27(5), 1413-1432.
39. Wang, J., Koseki, S., Chung, M.-J., Oh, D.-H., (2017). A novel approach to predict the growth of Staphylococcus aureus on rice cake. Frontiers in Microbiology 8(1140).
40. Xu, S., (2020a). Frequentist vs. Bayesian approaches in machine learning: A comparison of linear regression and bayesian linear regression. Medium.
41. Xu, S., (2020b). Generative vs. Discriminative probabilistic graphical models. A comparison of naive bayes and logistic regression. Medium.
42. Yamamoto, K., (2017). Food processing by high hydrostatic pressure. Bioscience, Biotechnology, and Biochemistry 81(4), 672-679.
43. Zhang, Z.-H., Wang, L.-H., Zeng, X.-A., Han, Z., Brennan, C.S., (2019). Non-thermal technologies and its current and future application in the food industry: a review. International Journal of Food Science & Technology 54(1), 1-13.