Drug response prediction model using a hierarchical structural component modeling method.

Area of Science:

• Biomarker discovery
• Proteomics
• Structural Equation Modeling

Background:

• Component-based structural equation modeling (SEM) is widely applied across scientific disciplines.
• This study applies SEM to biologically structured data, specifically proteomic peptide-level data from liver cancer patients.
• Multiple reaction monitoring mass spectrometry (MRM-MS) was used for precise peptide quantitation.

Purpose of the Study:

• To develop a novel component-based model for predicting drug response in liver cancer.
• To identify candidate drug-response biomarkers by integrating peptide and protein-level data.
• To facilitate biological interpretation of complex proteomic datasets.

Main Methods:

• Applied component-based structural equation modeling to proteomic peptide-level data.
• Utilized an alternating least squares algorithm for efficient estimation of peptide and protein coefficients.
• Employed permutation testing to select significant protein biomarkers, considering variable correlations.

Main Results:

• Developed a drug response prediction model that collapses peptide data into protein-level information.
• Successfully predicted liver cancer response to sorafenib, a tyrosine kinase inhibitor.
• Identified significant protein biomarkers for drug response prediction.

Conclusions:

• The developed model accurately predicts drug responses in liver cancer patients, achieving a high Area Under the Curve (AUC) score.
• This approach holds potential for clinical translation in identifying patients likely to benefit from specific therapies.
• Component-based SEM provides a robust framework for integrating and interpreting complex proteomic data for biomarker discovery.