LPE

LPE computes local-pooled-error estimates to perform significance analysis of microarray differential gene expression, improving variance estimation and detection of differential expression when the number of replicate arrays is limited.

Key Features:

• Resampling-based FDR adjustment: LPE employs resampling techniques for False Discovery Rate (FDR) adjustment, providing less conservative control than Benjamini-Hochberg (BH) or Benjamini-Yekutieli (BY) procedures.
• Local-pooled-error estimation: LPE pools errors within genes and between replicate arrays for genes with similar expression values to produce more reliable variance estimates with small sample sizes.
• Robust statistical tests: LPE applies statistical tests that leverage LPE variance estimates to assess significance of differential expression and reduce false positives from underpowered analyses.
• Data compatibility: Accepts raw text-format microarray data from MAS4, MAS5, and dChip platforms and also accepts normalized microarray data.
• Intensity-dependent error handling: LPE addresses heterogeneous error variability across biological conditions and intensity ranges that can make fold-change misleading.

Scientific Applications:

• Microarray differential-expression analysis with limited replicates: LPE is applied to identify significant differential gene expression when the number of replicate arrays is small.
• CD8+ T-cell activation studies: LPE has been used to compare gene expression between naïve and activated CD8+ T-cells, detecting differential-expression patterns with few replicates.

Methodology:

LPE computes local-pooled-error estimates by pooling errors within genes and between replicate arrays for genes with similar expression values, employs resampling-based FDR adjustment, applies robust statistical tests leveraging LPE estimates, and is implemented using S-PLUS and R functions.