Functional and Parametric Estimation in a Semi- and Nonparametric Model with Application to Mass-Spectrometry Data

Area of Science:

• Biostatistics
• Computational Biology
• Analytical Chemistry

Background:

• Mass spectrometry generates complex data requiring advanced statistical models.
• Existing methods may not fully capture both individual variations and common patterns in spectral data.

Purpose of the Study:

• To develop a novel semi- and nonparametric statistical model for mass spectrometry data.
• To simultaneously estimate parametric (location, scale) and nonparametric (shape) components.
• To validate the model's performance using simulations and real-world data.

Main Methods:

• Proposed a model with linear parametric components and a nonparametric regression function.
• Developed a multi-step estimation approach for simultaneous parameter and function estimation.
• Investigated theoretical properties including consistency and asymptotic normality for parametric parts.

Main Results:

• The proposed method achieves optimal convergence rates for the nonparametric component.
• Simulation studies confirm the method's effectiveness and good finite-sample performance.
• The model was successfully applied to SELDI-TOF mass spectrometry data from liver cancer patients.

Conclusions:

• The novel semi- and nonparametric model offers a robust approach for mass spectrometry data analysis.
• The method provides accurate estimation for both parametric and nonparametric features.
• This approach has practical implications for biomarker discovery in diseases like liver cancer.