Predictive comparison and evaluation of ANN and ANFIS as effective tools in modeling prodigiosin production by Serratia marcescens ND93
- 1Faculty of Engineering, Department of Electrical and Electronics Engineering, Ataturk University, Erzurum, Turkey
- 2Faculty of Agriculture, Department of Agricultural Biotechnology, Ataturk University, Erzurum, Turkey
- 3Faculty of Agricultural Sciences and Technology, Department of Field Crops, Sivas Science and Technology University, Sivas, Turkey
Abstract
Prodigiosin is a natural red pigment produced by Serratia marcescens that has attracted attention due to its various biological activities and potential for industrial applications. In recent years, research aimed at producing this compound using highly efficient and economically viable bioprocesses has gained increasing importance. In this study, the OD600 value and prodigiosin production of Serratia marcescens isolated from onion were experimentally tested under varying environmental conditions (glycerol, pH, temperature, and time), and artificial neural network (ANN) and ANFIS models were designed to predict prodigiosin production using the obtained data. First, the red pigment extracted with acidified methanol was characterized using a UV-vis spectrophotometer and exhibited maximum absorbance at 535 nm. Next, the modeling phase began, and both the ANN and ANFIS architectures were trained using 80% of the experimental dataset. The reserved 20% of the data served as the testing subset, encompassing input feature vectors including glycerol, pH, temperature, and time alongside their corresponding OD600 values and prodigiosin amount from the conventional system. Model performance validation was conducted using the Coefficient of Determination R² as the primary metric for assessing predictive accuracy. The ANN model demonstrated superior performance, achieving R² values of 0.9975 and 0.9942 on the training and testing sets, respectively. In contrast, the ANFIS model yielded significantly lower R² metrics of 0.9537 training and 0.9409 testing, respectively. The high R² values associated with the ANN technique signify a strong correlation between the model’s predicted outputs and the actual experimental observations. Although the ANFIS R² metrics are lower, they still indicate acceptable predictive error tolerances, particularly considering the constrained size of the experimental dataset. In summary, both ANN and ANFIS models successfully demonstrated efficacy in predicting both OD600 and prodigiosin amount; however, the ANN model consistently delivered a superior predictive capacity, exhibiting closer agreement with the experimental data. The findings of this study may help to better understand the application of ANN and ANFIS in modeling microbial pigment production and predicting production performance.
Supplementary Information: The online version contains supplementary material available at https://doi.org/10.1186/s12896-026-01210-5.
Related
- Caenorhabditis elegans and Serratia marcescens host-parasite coevolution model strains derived through mutagenesis and experimental evolution
- Comparison of Wildtype and Recombinant Adeno-Associated Virus (AAV) Production Reveals Distinct Productivity and Content Ratio Profiles.
- Identification of Microalgae, Evaluation of Water Purification Efficiency, and Optimization of Large-Scale Production Conditions Based on Recirculating Aquaculture Systems
- Phage therapy for multidrug-resistant Serratia marcescens infection: mechanisms, preclinical evidence, and translational challenges
- In vitro evaluation of bulk and nanosized cumin (Cuminum cyminum) extract on rumen gas production kinetics, methane and carbon dioxide productions, nutrient degradability, and fermentation end-products
BibTeX
@article{erenturk2026predictive,
title = {{Predictive comparison and evaluation of ANN and ANFIS as effective tools in modeling prodigiosin production by Serratia marcescens ND93}},
author = {Köksal Erentürk and Neslihan Dikbaş and Sevda Uçar and Şeyma Alım},
journal = {BMC Biotechnology},
year = {2026},
month = aug,
doi = {10.1186/s12896-026-01210-5},
url = {https://doi.org/10.1186/s12896-026-01210-5},
note = {CC BY-NC-ND 4.0},
}
RIS
TY - JOUR
TI - Predictive comparison and evaluation of ANN and ANFIS as effective tools in modeling prodigiosin production by Serratia marcescens ND93
AU - Köksal Erentürk
AU - Neslihan Dikbaş
AU - Sevda Uçar
AU - Şeyma Alım
PY - 2026
DA - 2026/08/07
JO - BMC Biotechnology
DO - 10.1186/s12896-026-01210-5
UR - https://doi.org/10.1186/s12896-026-01210-5
AB - Prodigiosin is a natural red pigment produced by Serratia marcescens that has attracted attention due to its various biological activities and potential for industrial applications. In recent years, research aimed at producing this compound using highly efficient and economically viable bioprocesses has gained increasing importance. In this study, the OD600 value and prodigiosin production of Serratia marcescens isolated from onion were experimentally tested under varying environmental conditions (glycerol, pH, temperature, and time), and artificial neural network (ANN) and ANFIS models were designed to predict prodigiosin production using the obtained data. First, the red pigment extracted with acidified methanol was characterized using a UV-vis spectrophotometer and exhibited maximum absorbance at 535 nm. Next, the modeling phase began, and both the ANN and ANFIS architectures were trained using 80% of the experimental dataset. The reserved 20% of the data served as the testing subset, encompassing input feature vectors including glycerol, pH, temperature, and time alongside their corresponding OD600 values and prodigiosin amount from the conventional system. Model performance validation was conducted using the Coefficient of Determination R² as the primary metric for assessing predictive accuracy. The ANN model demonstrated superior performance, achieving R² values of 0.9975 and 0.9942 on the training and testing sets, respectively. In contrast, the ANFIS model yielded significantly lower R² metrics of 0.9537 training and 0.9409 testing, respectively. The high R² values associated with the ANN technique signify a strong correlation between the model’s predicted outputs and the actual experimental observations. Although the ANFIS R² metrics are lower, they still indicate acceptable predictive error tolerances, particularly considering the constrained size of the experimental dataset. In summary, both ANN and ANFIS models successfully demonstrated efficacy in predicting both OD600 and prodigiosin amount; however, the ANN model consistently delivered a superior predictive capacity, exhibiting closer agreement with the experimental data. The findings of this study may help to better understand the application of ANN and ANFIS in modeling microbial pigment production and predicting production performance. Supplementary Information: The online version contains supplementary material available at https://doi.org/10.1186/s12896-026-01210-5.
SN - 1472-6750
ER -