Kabli, MohammadYilmaz, Mustafa TahsinTaylan, OsmanKaya, YaseminIspirli, HumeyraBasahel, AbdulrahmanSagdic, Osman2024-10-042024-10-0420200360-83521879-0550https://doi.org/10.1016/j.cie.2020.106619http://hdl.handle.net/20.500.12403/2967Optimisation of exopolysaccharides (EPS) production in Lactic Acid Bacteria (LAB) is an important task as EPS production can be affected by different parameters. In this respect, this study aimed to characterise the structure of an EPS from Leuconstoc mesenteroides DL1 strain and to optimise the EPS production by determination of the effects of incubation time, sucrose concentration, incubation temperature and initial levan concentration (input parameters) using integrated ANNs (Artificial neural networks) and fuzzy modelling approaches. The characterisation of the EPS monomeric composition by HPLC analysis revealed that EPS DL1 was composed of glucose and fructose. The H-1 and C-13 NMR spectra of EPS DL1 also confirmed the glucan and fructan production. The effects of the input parameters on glucan and fructan production levels as output parameters by DL1 were optimised using neural network and fuzzy modelling tools. The fuzzy model was developed based on the recognition of basic elements of input-output parameters, and the power of ANNs used for system identification. A structural analysis was carried out to improve the flexibility of fuzzy model, and to design the unknown mappings of the input and output parameters more robustly. The parameters then were fine-tuned by qualitative reasoning to establish the relations of input output parameters using membership functions (MFs) and their intervals determination. A hybrid training algorithm was employed for parameter identification, MFs and their interval determination to obtain the fuzzy model. The model can predict the outcome parameters; glucan and fructan with high accuracy for the predetermined input parameters.eninfo:eu-repo/semantics/closedAccessEPS productionStructural characterisationLactic acid bacteria (LAB)OptimisationNeural networksFuzzy modellingArticle14810.1016/j.cie.2020.1066192-s2.0-85089225619Q1WOS:000574658100001Q1