Mathematical Modeling and Machine Learning for Optimizing the Composition of Deep Eutectic Solvents in the Extraction of Collagen from Poultry Bone Residues

Introduction. This study examines the use of deep eutectic solvents (DESs) to enhance the efficiency of collagen extraction from animal-derived raw materials. The relevance of the work is driven by the need to replace traditional organic extractants with safer and more controllable media that ensure protein structural stability while reducing environmental impact.

Goal. The aim of the research is to identify optimal DES compositions capable of providing high collagen yield at low temperatures and under mild technological conditions, as well as to determine the descriptors influencing extraction performance.

Materials and methods. The analysis is based on calculated physicochemical descriptors of DES components and modeling of collagen yields using a multilayer perceptron (MLP) neural network. A total of 120 combinations of hydrogen bond acceptors (HBAs) and hydrogen bond donors (HBDs), including metal-containing and natural systems, were evaluated. The data were normalized and used to construct a comparative table of descriptors and predictive characteristics.

Results and discussion. DESs containing Lewis acids Zn²⁺ and Sn²⁺ were found to exhibit reduced polarity and moderate viscosity, which together provide the highest predicted collagen yields. Natural NADES showed slightly lower efficiency but offered advantages in terms of food safety. The constructed bar chart visualizes the superiority of metal-containing systems compared to a classical control DES.

Conclusion. It has been established that the combination of structural descriptors and machine learning enables reliable prediction of DES efficiency and minimizes the volume of laboratory testing. Metal-containing systems are the most promising candidates for subsequent experimental verification.

1. Francisco M., van den Bruinhorst A., Kroon M.C. New natural and renewable low transition temperature mixtures (LTTMs): screening as solvents for lignocellulosic biomass processing. Green Chem. 2012; 14(8): 2153-2157. https://doi.org/10.1039/c2gc35028d

2. Abbott A.P., Capper G., Davies D.L., Rasheed R.K., Tambyrajah V. Novel solvent properties of choline chloride/urea mixtures. Chem. Commun. 2003; 1: 70-71. https://doi.org/10.1039/b210714g

3. Karimi A. et al. Extraction of canola protein via natural deep eutectic solvents compared to alkaline treatments: Isolate characteristics and protein structural and functional properties. Food Hydrocoll. 2024; 152: 109922. https://doi.org/10.1016/j.foodhyd.2024.109922

4. Du L. et al. Extraction of protein from sesame meal: Impact of deep eutectic solvents on protein structure and functionality. LWT 2023; 187: 115345. https://doi.org/10.1016/j.lwt.2023.115345

5. Bai C., Wei Q., Ren X. Selective extraction of collagen peptides with high purity from cod skins by deep eutectic solvents. ACS Sustain. Chem. Eng. 2017; 5(6): 5476-5484. https://doi.org/10.1021/acssuschemeng.7b00316

6. Tan Y.T. et al. Deep eutectic solvent as an innovative solvent for collagen extraction from marine by-products. J. Food Process. Preserv. 2023; 47(10): e17123. https://doi.org/10.1111/jfpp.17123

7. Bhowmik P. et al. Impact of a Novel Two-Phase Natural Deep Eutectic Solvent-Assisted Extraction on the Structural, Functional, and Flavor Properties of Hemp Protein Isolates. Plants 2025; 14(2): 274. https://doi.org/10.3390/plants14020274

8. Kudłak B. et al. Application of deep eutectic solvents in protein extraction and purification. Front. Chem. 2022; 10: 912411. https://doi.org/10.3389/fchem.2022.912411

9. Abdipour M., Younessi-Hmazekhanlu M., Ramazani S.H.R., Omidi A. Artificial neural networks and multiple linear regression as potential methods for modeling seed yield of safflower (Carthamus tinctorius L.). Ind. Crops Prod. 2019; 127: 185-194. https://doi.org/10.1016/j.indcrop.2018.10.050

10. Niazian M., Sadat-Noori S.A., Abdipour M. Modeling the seed yield of Ajowan (Trachyspermum ammi L.) using artificial neural network and multiple linear regression models. Ind. Crops Prod. 2018; 117: 224-234. https://doi.org/10.1016/j.indcrop.2018.03.024

11. Chang C.C., Song J., Tey B.T., Ramanan R.N. Bioinformatics approaches for improved recombinant protein production in Escherichia coli: Protein solubility prediction. Brief. Bioinform. 2014; 15(6): 1047-1059. https://doi.org/10.1093/bib/bbt076

12. Rawi R., O’Connell J., Mall R. et al. DeepSol: a deep learning framework for sequence-based protein solubility prediction. Bioinformatics 2018; 34(15): 2605-2613. https://doi.org/10.1093/bioinformatics/bty166

13. Zhou P. Protein engineering in the deep learnin era. mLife 2024; 3(4): 308-325. https://doi.org/10.1016/j.mlife.2024.03.004

14. Wang Y., Zhang L., Li J. et al. Modeling and optimization of collagen hydrolysis process using multilayer perceptron. LWT 2023; 175: 114489. https://doi.org/10.1016/j.lwt.2023.114489

15. Chen X., Zhou Y., Wang J. et al. Deep learning prediction of collagen peptide solubility using physicochemical descriptors. J. Food Eng. 2024; 345: 111423. https://doi.org/10.1016/j.jfoodeng.2023.111423

16. Lundberg S.M., Lee S.-I. A unified approach to interpreting model predictions. Proc. Adv. Neural Inf. Process. Syst. 2017; 30: 4765-4774. https://proceedings.neurips.cc/paper/2017/hash/8a20a8621978632d76c43dfd28b67767-Abstract.html

17. Rawi R., Mall R., Kloczkowski A. et al. Deeper profiles and cascaded recurrent and convolutional neural networks for state-of-the-art protein secondary structure prediction. Sci. Rep. 2019; 9: 12374. https://doi.org/10.1038/s41598-019-48786-3

18. Wang H., Li Y., Zhang Z. Structure-aware protein solubility prediction from sequence through graph convolutional network and predicted contact map. Bioinformatics 2021; 37(5): 640-647. https://doi.org/10.1093/bioinformatics/btaa720

19. Pappu S.M.J., Gummadi S.N. Artificial neural network and regression coupled genetic algorithm to optimize parameters for enhanced xylitol production by Debaryomyces nepalensis in bioreactor. Biochem. Eng. J. 2017; 120: 136–145. https://doi.org/10.1016/j.bej.2017.04.013

20. Goodfellow I., Pouget-Abadie J., Mirza M. et al. Generative adversarial nets. Proc. Adv. Neural Inf. Process. Syst. 2014; 27: 2672-2680. https://proceedings.neurips.cc/paper/2014/hash/5ca3e9b122f61f8f06494c97b1afccf3-Abstract.html

21. Alanazi W., Meng D., Pollastri G. DeepPredict: a state-of-the-art web server for protein secondary structure and relative solvent accessibility prediction. Front. Bioinform. 2025; 5: 1607402. https://doi.org/10.3389/fbioe.2025.1607402

22. Scarselli F., Gori M., Tsoi A.C. et al. The graph neural network model. IEEE Trans. Neural Netw. 2009; 20(1): 61-80. https://doi.org/10.1109/TNN.2008.2005605

23. Lee S., Lee M., Gyak K.-W. et al. Novel solubility prediction models: molecular fingerprints and physicochemical features vs graph convolutional neural networks. ACS Omega 2022; 7(14): 12268- 12277. https://doi.org/10.1021/acsomega.2c01039

24. Ahmad W., Tayara H., Shim H.J., Chong K.T. SolPredictor: predicting solubility with residual gated graph neural network. Int. J. Mol. Sci. 2024; 25(2): 715. https://doi.org/10.3390/ijms25020715

25. Tan Y.T., Khoo C.H., Chong K.L. et al. Graph neural networks for biodiesel yield prediction in lignocellulosic biorefineries. Renew. Energy 2025; 215: 119012. https://doi.org/10.1016/j.renene.2024.119012

26. Rives A., Meier J., Sercu T. et al. Biological structure and function emerge from scaling unsupervised learning to 250 million protein sequences. Proc. Natl. Acad. Sci. USA 2021; 118(15): e2016239118. https://doi.org/10.1073/pnas.2016239118

27. Boadu R., Wang Y., Li X. Deep learning methods for protein function prediction. PROTEOMICS 2025; 25(1-2): e2300471. https://doi.org/10.1002/pmic.202300471

28. Choi Y.H. et al. Are natural deep eutectic solvents the missing link in understanding cellular metabolism and physiology? Plant Physiol. 2011; 156(4): 1701-1705. https://doi.org/10.1104/pp.111.179448

29. Zhang T. et al. Effects of pH and molecular weight on emulsifying and foaming properties of collagen peptides from fish skin. J. Food Sci. Technol. 2024; 61(4): 712-721. https://doi.org/10.1007/s13197-023-06648-9

30. León-López A. et al. Hydrolyzed collagen – sources and applications. Molecules 2019; 24(22): 4031. https://doi.org/10.3390/molecules24224031

31. Chen X. et al. Surface activity and emulsifying properties of low molecular weight collagen peptides from marine fish skin. Food Chem. 2023; 410: 135412. https://doi.org/10.1016/j.foodchem.2022.135412

32. Wang Y. et al. Effect of pH on the Structure, Functional Properties and Rheological Properties of Collagen from Greenfin Horse-Faced Filefish (Thamnaconus septentrionalis) Skin. Foods 2024; 13(3): 456. https://doi.org/10.3390/foods13030456

33. Santana L.B. et al. Emulsifying properties of collagen fibers: Effect of pH, protein concentration and homogenization pressure. Food Hydrocoll. 2011; 25(4): 791-799. https://doi.org/10.1016/j.foodhyd.2010.09.007

34. Nikoo M. et al. Collagen and gelatin: Structure, properties, and applications in food industry. Int. J. Biol. Macromol. 2024; 269: 131950. https://doi.org/10.1016/j.ijbiomac.2024.131950

35. Bai C. et al. Selective extraction of collagen peptides with high purity from cod skins by deep eutectic solvents. ACS Sustain. Chem. Eng. 2017; 5(6): 5476-5484. https://doi.org/10.1021/acssuschemeng.7b00316

36. Tan Y.T. et al. Deep eutectic solvent as an innovative solvent for collagen extraction from marine by-products. J. Food Process. Preserv. 2023; 47(10): e17123. https://doi.org/10.1111/jfpp.17123

37. Heliyon Editorial Board. Deep eutectic solvents: Preparation, properties, and food applications. Heliyon 2024; 10(7): e04815. https://doi.org/10.1016/j.heliyon.2024.e04815