LSPR

LSPR detects periodic expression profiles in DNA microarray time-series data using Lomb-Scargle periodogram and harmonic regression to characterize cyclic gene expression for studies of circadian rhythms and other temporal biological processes.

Key Features:

• Preprocessing: Removes linear trend from raw time-series data and filters noise to emphasize periodic components.
• Lomb-Scargle Periodogram: Uses Lomb-Scargle periodogram to detect and estimate frequencies of periodic signals in unevenly sampled data.
• Harmonic Regression Modeling: Applies harmonic regression to model detected cyclic components and characterize amplitude and phase.
• False Discovery Rate Procedure: Implements an FDR procedure to select inferred periodic transcripts and control false positives.
• MATLAB implementation: Provided as a MATLAB package for execution of the analysis.
• Benchmarking: Tested on synthetic unevenly sampled datasets and two Arabidopsis diurnal expression datasets and compared to established algorithms.

Scientific Applications:

• Circadian and temporal gene expression analysis: Identification and characterization of genes exhibiting cyclic expression driven by circadian rhythms or other time-dependent processes.
• Detection of periodic transcripts in microarray time-series: Selection of transcripts with statistically supported periodicity from DNA microarray experiments.
• Genomics and systems biology studies: Analysis of temporal regulation and timing of gene expression to inform systems-level models.
• Method benchmarking: Evaluation of periodicity detection performance on synthetic and Arabidopsis diurnal datasets against existing algorithms.

Methodology:

Preprocessing with linear-trend removal and noise filtering; Lomb-Scargle periodogram for frequency detection in unevenly sampled time-series; harmonic regression to model amplitude and phase of cyclic components; false discovery rate procedure to select periodic transcripts; implemented in MATLAB.