Spectral estimation in unevenly sampled space of periodically expressed microarray time series data

Area of Science:

• Genomics
• Bioinformatics
• Signal Processing

Background:

• Microarray time-series analysis is crucial for identifying periodically expressed genes.
• Challenges include noise, short length, and uneven sampling in biological data.
• Existing methods often fail with unevenly sampled time-series data.

Purpose of the Study:

• To develop a novel spectral estimation algorithm for unevenly sampled gene expression data.
• To improve the detection of periodic genes in microarray time-series.
• To provide a tool for gene expression time-series interpolation or resampling.

Main Methods:

• A new spectral estimation algorithm based on signal reconstruction in a shift-invariant signal space.
• Utilizes B-spline basis for direct spectral estimation.
• Evaluated on simulated and real gene expression data (Plasmodium falciparum, Yeast).

Main Results:

• The proposed algorithm demonstrates superior performance in detecting periodically expressed genes compared to Lomb-Scargle and Fourier periodogram methods.
• Successfully identified biologically meaningful periodic genes in Plasmodium falciparum and Yeast datasets.
• The method is robust to noise and data imperfections.

Conclusions:

• An effective method for identifying periodic genes in unevenly sampled microarray time-series data has been developed.
• The algorithm offers a valuable tool for gene expression time-series interpolation and resampling.
• The approach enhances the analysis of biological rhythms in gene expression.