Strategy for reduced calibration sets to develop quantitative structure-retention relationships in high-performance...

Area of Science:

• Analytical Chemistry
• Computational Chemistry
• Chromatography

Background:

• Quantitative structure-retention relationships (QSRRs) model analyte retention in chromatography based on molecular descriptors.
• Traditional QSRR model development requires large numbers of analytes, making it time-consuming and laborious.
• Developing efficient QSRR models is crucial for accelerating chemical analysis and prediction.

Purpose of the Study:

• To present a strategy for building QSRR models using reduced calibration sets.
• To demonstrate the effectiveness of the Kennard and Stone algorithm for selecting analytes in reduced sets.
• To establish guidelines for constructing small, representative calibration sets for QSRR modeling.

Main Methods:

• Utilized the Kennard and Stone algorithm to select analytes for reduced calibration sets based on molecular descriptors.
• Applied the strategy to develop and validate three QSRR models of varying complexity (logP, quantum chemical indices, LSER descriptors).
• Tested the models across 76 reversed-phase high-performance liquid chromatography systems.

Main Results:

• Achieved successful QSRR model development for logP with as few as seven analytes.
• Determined that three analytes per descriptor are sufficient for quantum chemical indices (QCI) and linear solvation energy relationship (LSER) models.
• Confirmed that reduced calibration sets adequately cover both descriptor and retention variable spaces.

Conclusions:

• Reduced calibration sets significantly streamline QSRR model development in chromatography.
• The Kennard and Stone algorithm is effective for selecting optimal analytes for small calibration sets.
• Guidelines are provided for constructing efficient QSRR calibration sets, reducing experimental effort.