Machine Learning Approach for Determining the Formation of β-Lactam Antibiotic Complexes with Cyclodextrins Using Multispectral Analysis

The problem of determining the formation of complexes of β-lactam antibiotics with cyclodextrins (CDs) and the interactions involved in this process were addressed by machine learning on multispectral images. Complexes of β-lactam antibiotics, including cefuroxime axetil, cefetamet pivoxil, and pivampicillin, as well as CDs, including αCD, βCD, γCD, hydroxypropyl-αCD, methyl-βCD, hydroxypropyl-βCD, and hydroxypropyl-γCD, were prepared in all combinations. Thermograms confirming the formation of cyclodextrin complexes were obtained using differential scanning calorimetry. Transmission Fourier-transform infrared (tFTIR) and complementary attenuated total reflectance FTIR (ATR) coupled with machine learning were techniques chosen as a nondestructive alternative. The machine learning algorithm was used to determine the formation of complexes in samples using solely their tFTIR and ATR spectra at the prediction stage. Parameterized method 7 (PM7) was used to support the analysis by molecular modeling of the complexes. The model developed through machine learning properly distinguished samples with formed complexes form noncomplexed samples with a cross-validation accuracy of 90.4%. Analysis of the contribution of spectral bands to the model indicated interactions of ester groups of β-lactam antibiotics with CDs, as well as some interactions of cephem ring in cefetamet pivoxil and penam moiety in pivampicillin. Molecular modeling with PM7 helped to explain experimental results and allowed to propose possible binding modes.