Evaluation of ChIPXpress performance across different truncation cutoffs for a set ranking weight of 0.1. Boxplot of the performance of different ranking methods for GEO-only, ChIPx-only, and ChIPXpress using TF expression cutoffs of c = −∞ (All), -2, -1, 0, 1, 2. The ranking weight is set to w = 0.1 for each method tested. Performance is measured by the normalized AUC score (the area under the positive predictive value curve divided by the total plot area). This is equivalent to averaging the positive predictive values across the top 1 to n (100, 500, 1000) ranked predictions. Each individual point corresponds to a specific TF data set and the crosses correspond to the mean normalized AUC score across all datasets tested. CisGenome was used to analyze ChIP-seq data and TileProbe was used to analyze ChIP-chip data.