In previous validation results, we assessed clustering by examining the ranking of true positives. Here we assess how well the fold change between experimental conditions is recovered. For each contig matching a gene with true differential expression, we compared its cluster-level log2 fold change against its true gene-level log2 fold change. The Pearson correlation between these quantities is shown. We assessed each clustering method in this way and found corset clustering gave the highest correlation in all cases. The highest Pearson correlation for each assembly is displayed in bold.