Meta-analysis for pathway enrichment analysis when combining multiple genomic studies

Abstract

Motivation: Many pathway analysis (or gene set enrichment analysis) methods have been developed to identify enriched pathways under different biological states within a genomic study. As more and more microarray datasets accumulate, meta-analysis methods have also been developed to integrate information among multiple studies. Currently, most meta-analysis methods for combining genomic studies focus on biomarker detection and meta-analysis for pathway analysis has not been systematically pursued.

Results: We investigated two approaches of meta-analysis for pathway enrichment (MAPE) by combining statistical significance across studies at the gene level (MAPE_G) or at the pathway level (MAPE_P). Simulation results showed increased statistical power of meta-analysis approaches compared to a single study analysis and showed complementary advantages of MAPE_G and MAPE_P under different scenarios. We also developed an integrated method (MAPE_I) that incorporates advantages of both approaches. Comprehensive simulations and applications to real data on drug response of breast cancer cell lines and lung cancer tissues were evaluated to compare the performance of three MAPE variations. MAPE_P has the advantage of not requiring gene matching across studies. When MAPE_G and MAPE_P show complementary advantages, the hybrid version of MAPE_I is generally recommended.

Availability: http://www.biostat.pitt.edu/bioinfo/

Contact: ctseng@pitt.edu

Supplementary information: Supplementary data are available at Bioinformatics online.

Fig. 1.

Diagram for MAPE analysis. (A) Pathway enrichment analysis for an individual study; (B) MAPE_G; (C) MAPE_P; (D): MAPE_I.

Fig. 2.

(A and B) Heatmaps of genes in two example pathways identified by MAPE_P alone and by MAPE_G alone, respectively, in lung cancer studies. AANU in 2A is detected by MAPE_P (q = 0.007) but not by MAPE_G (q = 0.073) whereas HCTU in 2B is detected by MAPE_G (q = 0.016) but not MAPE_P (q = 0.071). The Q-values of each individual gene (on the row) and study (on the column) are shown by gradient color in −log (base 10) scale. Gene set I contains biomarkers with Q-value <0.05 for both studies and gene sets II and III contain significant biomarkers (q < 0.05) in one of the studies. (C and D) Venn diagram of biomarkers detected by each individual study (Beer and Bhat) in AANU and HCTU. (E) Power difference between MAPE_P and MAPE_G for various α and λ when θ₁ = θ₂ = 0.5, 0.75, 1, 1.5, 2 and 4. Yellow/red color shows better power of MAPE_P over MAPE_G and blue color vice versa. Solid lines show contours of equal power between MAPE_P and MAPE_G. (F) The blue, green and red lines indicates the power of MAPE_P, MAPE_G and MAPE_I, respectively, when θ = 4, α = 0.2. (G) Power difference between MAPE_P and MAPE_G for various α and λ when combining 2, 4 or 10 studies.

Fig. 3.

Detailed algorithms of MAPE_G, MAPE_P and MAPE_I.

Fig. 4.

Venn diagram comparing pathways detected by individual studies and by the three MAPE methods. (A and B) Application in combining Liedtke and Neve breast cancer drug response studies. (C and D) Application in combining Beer and Bhat lung cancer studies.