Quantile Index Biomarkers Based on Single-Cell Expression Data

Area of Science:

• Biomedical science
• Computational pathology
• Cancer research

Background:

• Histocytometry enables single-cell biomolecule quantification in tumors.
• Current methods often ignore expression heterogeneity, using only mean signal intensity.
• This overlooks valuable information within the cellular signal intensity (CSI) distribution.

Purpose of the Study:

• To develop a novel biomarker utilizing the entire CSI distribution for clinical outcome prediction.
• To assess the prognostic value of this new biomarker in malignant breast tumors.
• To provide a computational tool for implementing the new biomarker.

Main Methods:

• Defined a quantile index (QI) biomarker as a weighted average of CSI distribution quantiles.
• Optimized quantile weights using functional regression models for clinical outcome prediction.
• Applied QI biomarkers to protein expression data from breast cancer patients.

Main Results:

• QI biomarkers demonstrated improved prognostic value over standard mean signal intensity predictors.
• The method effectively captures and utilizes expression heterogeneity for prediction.
• An R package, Qindex, was developed for QI biomarker implementation.

Conclusions:

• The quantile index (QI) offers a more powerful approach to cancer biomarker development by considering expression distribution.
• This method enhances prediction of clinical outcomes in breast cancer.
• The approach is versatile and applicable to various cell-level expression data beyond immunohistochemistry.