Measuring the consistency of the TP53→CDKN1A link across independent studies.
Single studies can be misleading. A meta-analysis pools results across independent sources to test whether an effect holds across tissues and labs.
The forest plot visualizes meta-analytic results. Each row is a study with an effect size and confidence interval. The square marks the point estimate, and its area scales with the weight assigned to that study. The diamond at the bottom is the pooled estimate.
Figure 1. Eight studies reporting CDKN1A log₂ fold change. Squares scale with study weight; dashed line is the null. Switch the pooling model to see how between-study heterogeneity changes the weights and the pooled diamond.
The I² statistic estimates what fraction of the observed variation is due to between-study heterogeneity rather than sampling error. It is not a biological proof by itself. High I² asks you to explain why studies differ: tissue context, assay platform, cohort composition, or analysis choices can all matter.
Leave-one-out analysis removes each study in turn to test whether any single observation is driving the conclusion.
Figure 2. Leave-one-out explorer. The diamond and I² statistic shift as studies are excluded. A robust effect should persist after removing the largest study.
Journals tend to prefer significant results, which can bias the literature. The funnel plot plots effect size against standard error. In a large, unbiased collection, smaller studies scatter more widely and roughly symmetrically around the pooled estimate. With only a few studies, asymmetry is a hint to investigate, not a diagnosis.
Figure 3. Funnel plot for detecting bias. The dashed line is the pooled estimate; asymmetry suggests missing studies — often smaller ones with non-significant findings.