Figure 8.

A comparison of each method in 1v1 subsampling of a Duchenne muscular dystrophy dataset. The most positive results are clearly seen in the two voting methods combining the GEO results with either the local z-score method (as implemented in SNOMAD) or the percentage difference method (using percent changes relative to the mean expression levels). These methods return 96% more of the top 50 genes than the standard method alone and 80% more of the top 50 genes within 100 genes.