Retention Index Prediction Using Quantitative Structure-Retention Relationships for Improving Structure...

Area of Science:

• Analytical Chemistry
• Computational Chemistry
• Metabolomics

Background:

• Nontargeted metabolomics using liquid-chromatography coupled to mass spectrometry (LC-MS) faces challenges in structure identification.
• Quantitative structure-retention relationship (QSRR) modeling can predict metabolite retention times, aiding in false positive elimination.

Purpose of the Study:

• To develop and validate a QSRR-based dual-filtering strategy for reducing false positives in nontargeted metabolomics.
• To improve the accuracy of metabolite structure identification in complex biological samples.

Main Methods:

• Compounds were grouped by molecular weight, and QSRR models were built using partial least squares (PLS) regression and a Genetic Algorithm (GA) with VolSurf+ descriptors.
• A dual-filtering approach combining Tanimoto similarity (TS) and retention index (RI) similarity clustering was employed.
• QSRR models were trained on selected compounds to predict retention behavior.

Main Results:

• QSRR models achieved an R² of 0.8512 and an average RMSEP of 8.45%.
• The dual-filtering approach predicted representative compounds with >91% accuracy.
• False positives were eliminated for 53% of the compound groups analyzed (18/34).

Conclusions:

• The proposed dual-filtering strategy effectively reduces false positives in nontargeted metabolomics.
• This method accelerates structure identification and improves the reliability of metabolomics data interpretation.
• The approach offers a robust solution for enhancing the efficiency of LC-MS-based metabolomics studies.