The present study aimed to explore the application of machine learning-based tools to study the impact of physicochemical properties of peptides and filtration membranes (FM) on peptide migration during electrodialysis with filtration membranes (EDFM). A total of 14 membranes characterized in terms of 10 physicochemical properties were employed to evaluate the selective migration of cationic and anionic peptides from a tryptic hydrolysate of whey protein isolate well characterized by chemometric and bioinformatic methods. Two machine learning approaches were compared: Decision Tree (DT) and Binary Greedy Network (BGN). Based on the feature selection and model performance results, DT appeared to be the most appropriate approach to generate explanatory models of peptide migration. From the selected DT models, selective migration patterns associated with specific physicochemical properties of peptides and FMs were established for the first time. The migration of anionic peptides was positively affected by higher average peak-to-valley roughness (Rz) values and macropores distribution in the filtrating layer (Mp-FL), where S11N+ and PES300 were the FMs providing the highest recovery of this type of peptides, particularly for those with isoelectric point (pI) values lower than 4.407 (IDALNENK, KYLLFCMENSAEPEQSLACQCLVRTPEVD, SLAMAASDISLLDAQSAPL, SLAMAASDISLLDAQSAPLR, TDYKKYLLFCMENSAEPEQ, TPEVDDEALEK, TPEVDDEALEKFDK, VLVLDTDYK, and VYVEELKPTPEGDLEILLQK) which reached average recovery rates of 16.617 and 4.556%, respectively. Regarding the migration of cationic peptides, this was mainly affected by other membrane characteristics such as volumetric porosity (Vp) and zeta potential (ZP), as well as peptide parameters such as polar residue content, pI, and leucine content. S11, S11SO3-, PES100, and PES300 were the FMs with the highest selective recovery rate for 7 out of 11 cationic peptides (ALPHMIR, IPAVFK, PMHI, PMHIR, TKIPAVF, TKIPAVFK, and VGINYWLAHK) with average values between 2.149 to 4.328%. This novel DT-based approach represents a suitable tool for studying peptide migration, choosing the most appropriate FMs for selective bioactive peptide migration during EDFM, and understanding the phenomena involved in their migration process.
Read full abstract