Region-wise chemometric framework for identifying FTIR functional-group signatures associated with cholinesterase inhibition.

Two key therapeutic targets in the treatment of Alzheimer's disease are acetylcholinesterase (AChE) and butyrylcholinesterase (BuChE). Understanding the molecular characteristics associated with cholinesterase inhibition is important for the development of new inhibitors. In this study, a chemometric framework integrating Fourier Transform Infrared (FTIR) spectroscopy, Principal Component Analysis (PCA), and region-wise Partial Least Squares Regression (PLSR) was applied to explore spectral-response relationships in newly synthesized flurbiprofen hydrazide and hydrazone derivatives. The dataset comprised 27 compounds with 1868 spectral variables covering the 4000-400 cm-1 region and experimentally determined AChE and BuChE inhibition activities. Following SNV preprocessing, PCA was used to characterize spectral variability and define five major spectral domains: O-H/N-H, C-H, carbonyl/amide, C-O/C-N, and fingerprint regions. Region-wise PLSR models were evaluated using 5-fold cross-validation based on RMSECV and MAE-CV, with Y-randomization providing an additional assessment of the observed modeling performance. Among the investigated regions, the C-H stretching domain (3000-2800 cm-1) showed the most favorable relative cross-validated performance for both AChE (RMSECV = 5.49; MAE-CV = 4.02) and BuChE (RMSECV = 5.09; MAE-CV = 5.13). Y-randomization further supported the response-related performance of the C-H region, particularly for BuChE. Regression coefficient profiles provided additional spectral information for interpreting the response-related variation across the investigated regions. Overall, the FTIR-PCA-PLSR framework provides an interpretable approach for comparative assessment of spectral domains and for generating chemically testable hypotheses regarding cholinesterase inhibition in the investigated flurbiprofen derivative series.