Statistical multiple-testing correction · BEGINNER LESSON
BH correction
BH correction, short for Benjamini-Hochberg correction, is a multiple-testing method that adjusts many p-values together to control the false discovery rate (FDR). It is used when many hypotheses are tested at once, such as in variant studies, chromatin-loop analysis, or proteomics.
What it is and what is measured
BH correction is a way to adjust p-values when many tests are done together. A p-value is a number from a statistical test that summarizes how surprising the data are if there were no real effect. The main thing BH correction tries to control is the false discovery rate (FDR), meaning the expected fraction of declared positives that are actually false positives. In the sources, BH correction is used after p-values are computed for variant effects, chromatin interactions, or protein abundance changes.
[1] Accurate variant effect estimation in FACS-based deep ... - PMC [2] DiffHiChIP: Identifying differential chromatin contacts from ... [3] Ground Truth–Based Evaluation of False Discovery Rate and ... - PMC
How it works
The basic idea is to rank the p-values from smallest to largest and then adjust them so that the list is judged as a group instead of one test at a time. This makes it easier to limit the share of false discoveries among the results you call significant. In the cited studies, BH correction was applied to p-values from normal-model tests, Welch’s t-tests, edgeR models, and other workflows before deciding which findings to report.
[1] Accurate variant effect estimation in FACS-based deep ... - PMC [2] DiffHiChIP: Identifying differential chromatin contacts from ... [3] Ground Truth–Based Evaluation of False Discovery Rate and ... - PMC
Why researchers use it
BH correction is popular because it is simple and widely used for large datasets. In the examples, it was used to control FDR in deep mutational scanning, HiChIP loop detection, and DIA proteomics. It can give reasonable discovery sets and can work well when the p-values are well behaved and the experiment is not strongly affected by hidden sources of variation.
[1] Accurate variant effect estimation in FACS-based deep ... - PMC [2] DiffHiChIP: Identifying differential chromatin contacts from ... [3] Ground Truth–Based Evaluation of False Discovery Rate and ... - PMC
Limitations and caveats
BH correction does not repair problems in the data or the test itself. The sources show several limits: it does not account for replicate-to-replicate variance, read coverage differences, or distance-related effects in chromatin contacts. It can also miss long-range loops when lower counts weaken statistical power. In proteomics, nominal BH-FDR underestimated the true FDR, meaning the reported error rate was too optimistic because some error came from sample handling and measurement steps outside the final statistical model. So BH correction controls error only under the assumptions of the input p-values; it is not a direct measure of the real biological error rate.
[1] Accurate variant effect estimation in FACS-based deep ... - PMC [2] DiffHiChIP: Identifying differential chromatin contacts from ... [3] Ground Truth–Based Evaluation of False Discovery Rate and ... - PMC
SEE THE IDEA
How BH correction is used in a study
One experiment produces many p-values, one for each variant, loop, or protein.
The p-values are adjusted together so discoveries can be judged as a set.
The method aims to limit the expected fraction of false positives among reported hits.
Hidden variance, low counts, or poorly calibrated p-values can still make the result misleading.
Many test results go into BH adjustment, which is then used to control FDR, but the quality of the result still depends on the data and the statistical model.
BH correction organizes many p-values into one multiple-testing decision, yet it cannot fix every source of experimental error.
TRY IT YOURSELF
Test your understanding
Choose an answer, then check the explanation.
PUT IT TOGETHER
Match each concept to its meaning
Drag a label onto its description, or select a label and then select a description. This also works with a keyboard.
4 pairs to match
Sources and further reading
This general lesson was generated from public source material. It explains a concept; it does not interpret a private paper or provide medical advice.