Student or educator? Sign up with your school email and get 6 months of Pro free.
CoChat

The DOI matches its Crossref record; 54 of the 55 references it cites were found in a registry or an index.

Rank–rank hypergeometric overlap: identification of statistically significant overlap between gene-expression signatures

Open access

Abstract

Comparing independent high-throughput gene-expression experiments can generate hypotheses about which gene-expression programs are shared between particular biological processes. Current techniques to compare expression profiles typically involve choosing a fixed differential expression threshold to summarize results, potentially reducing sensitivity to small but concordant changes. We present a threshold-free algorithm called Rank-rank Hypergeometric Overlap (RRHO). This algorithm steps through two gene lists ranked by the degree of differential expression observed in two profiling experiments, successively measuring the statistical significance of the number of overlapping genes. The output is a graphical map that shows the strength, pattern and bounds of correlation between two expression profiles. To demonstrate RRHO sensitivity and dynamic range, we identified shared expression networks in cancer microarray profiles driving tumor progression, stem cell properties and response to targeted kinase inhibition. We demonstrate how RRHO can be used to determine which model system or drug treatment best reflects a particular biological or disease response. The threshold-free and graphical aspects of RRHO complement other rank-based approaches such as Gene Set Enrichment Analysis (GSEA), for which RRHO is a 2D analog. Rank-rank overlap analysis is a sensitive, robust and web-accessible method for detecting and visualizing overlap trends between two complete, continuous gene-expression profiles. A web-based implementation of RRHO can be accessed at http://systems.crump.ucla.edu/rankrank/.

Overview

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1093/nar/gkq636

Cites

54 of 54 references verified0 citation errors · 0 not checkable

Statistics

Figures

Figure 1. Schematic of the RRHO algorithm. (A) Preprocessing: gene-expression profiles are defined as continuous lists of genes ranked by their degree of differential expression. Here, we rank by the signed log10-transformed t-test P-values between Class A and B, positive sign if the mean expression is higher in Class B and negative if lower. Microarray probes measuring the same gene are collapsed to a single value using the most differentially expressed probe and only genes measured in both experiments are retained. A scatter plot can be used to visualize gene ranking similarities, particularly when the degree of overlap is high. The regression fit (red line) and Spearman’s rank correlation coefficient (ρ) for this example are indicated. Creating the RRHO map: the hypergeometric P-value for enrichment of k overlapping genes is calculated for all possible threshold pairs for each experiment, creating a matrix where the indices are the current rank in each experiment (s, M). The direction-signed log10-transformed hypergeometric P-values are plotted in a heatmap as indicated by an accompanying color scale, mapping the degree of statistically significant overlap between the two ranked lists from that point on the map (corresponding to a rank threshold pair) to either the bottom left (ranks 0,0) or top right corner (ranks N,N). Multiple hypothesis correction can be applied using either permutation analysis (when sample number is large) or a BY FDR correction. (B) Interpreting RRHO maps. (i) Different map patterns indicate different types of overlap, such as the full profiles being correlated or only genes increasing in both experiments overlapping. (ii) The highest intensity point on the map can be used to extract the most statistically significant overlapping gene set. RRHO analysis can be used (iii) to compare relative overlap pairwise within a set of profiles (e.g. Figures 3B and 4) or (iv) to compare an experimental profile to a series of reference signatures (e.g. Figure 5).
Figure 1. Schematic of the RRHO algorithm. (A) Preprocessing: gene-expression profiles are defined as continuous lists of genes ranked by their degree of differential expression. Here, we rank by the signed log10-transformed t-test P-values between Class A and B, positive sign if the mean expression is higher in Class B and negative if lower. Microarray probes measuring the same gene are collapsed to a single value using the most differentially expressed probe and only genes measured in both experiments are retained. A scatter plot can be used to visualize gene ranking similarities, particularly when the degree of overlap is high. The regression fit (red line) and Spearman’s rank correlation coefficient (ρ) for this example are indicated. Creating the RRHO map: the hypergeometric P-value for enrichment of k overlapping genes is calculated for all possible threshold pairs for each experiment, creating a matrix where the indices are the current rank in each experiment (s, M). The direction-signed log10-transformed hypergeometric P-values are plotted in a heatmap as indicated by an accompanying color scale, mapping the degree of statistically significant overlap between the two ranked lists from that point on the map (corresponding to a rank threshold pair) to either the bottom left (ranks 0,0) or top right corner (ranks N,N). Multiple hypothesis correction can be applied using either permutation analysis (when sample number is large) or a BY FDR correction. (B) Interpreting RRHO maps. (i) Different map patterns indicate different types of overlap, such as the full profiles being correlated or only genes increasing in both experiments overlapping. (ii) The highest intensity point on the map can be used to extract the most statistically significant overlapping gene set. RRHO analysis can be used (iii) to compare relative overlap pairwise within a set of profiles (e.g. Figures 3B and 4) or (iv) to compare an experimental profile to a series of reference signatures (e.g. Figure 5).CC BY-NC · PubMed Central
Figure 2. RRHO Maps are a sensitive method for detection and visualization of overlap in expression data. Three representations of overlapping published cancer-related gene-expression signatures: signatures with strong (A–C), medium (D–F) and weak (G–I) overlap represented as metric scatter plots (A, D and G), rank–rank scatter plots (B, E and H) and hypergeometric overlap maps (C, F and I). The first row of metric scatter plots show genes plotted by their direction-signed (up, positive; down, negative), log10-transformed t-test P-values in each experiment; genes significantly changing in the same direction in both experiments are in Cartesian quadrants I and III and in opposite directions in quadrants II and IV. The Spearman’s ρ and Kendall’s τ rank correlation coefficients are indicated. The second row of rank–rank scatter plots show each gene plotted by its rank based on this metric. This representation spreads the genes more evenly across the plot and allows for assessment of overlap by increases in local density. The resulting plots show a higher density of genes along the diagonal in both the strong and medium overlap cases, especially at the bottom left and top right regions. The last row shows the rank–rank hypergeometric heatmap for each of these comparisons, where the overlap is represented statistically based on the hypergeometric distribution allowing visualization of any signal, even those that are relatively weak but significant. The log10-transformed hypergeometric P-values are indicated in the color scale bar with negative values indicating under-enrichment. Sample permutation P-values based on the sum of the signal in the bottom left and top right regions are indicated (perm pval).
Figure 2. RRHO Maps are a sensitive method for detection and visualization of overlap in expression data. Three representations of overlapping published cancer-related gene-expression signatures: signatures with strong (A–C), medium (D–F) and weak (G–I) overlap represented as metric scatter plots (A, D and G), rank–rank scatter plots (B, E and H) and hypergeometric overlap maps (C, F and I). The first row of metric scatter plots show genes plotted by their direction-signed (up, positive; down, negative), log10-transformed t-test P-values in each experiment; genes significantly changing in the same direction in both experiments are in Cartesian quadrants I and III and in opposite directions in quadrants II and IV. The Spearman’s ρ and Kendall’s τ rank correlation coefficients are indicated. The second row of rank–rank scatter plots show each gene plotted by its rank based on this metric. This representation spreads the genes more evenly across the plot and allows for assessment of overlap by increases in local density. The resulting plots show a higher density of genes along the diagonal in both the strong and medium overlap cases, especially at the bottom left and top right regions. The last row shows the rank–rank hypergeometric heatmap for each of these comparisons, where the overlap is represented statistically based on the hypergeometric distribution allowing visualization of any signal, even those that are relatively weak but significant. The log10-transformed hypergeometric P-values are indicated in the color scale bar with negative values indicating under-enrichment. Sample permutation P-values based on the sum of the signal in the bottom left and top right regions are indicated (perm pval).CC BY-NC · PubMed Central
Figure 3. RRHO identifies statistically significant overlap between expression signatures supporting or generating biological hypotheses. (A) High overlap in gene-expression changes between mouse models of prostate neoplasia driven by PTEN and AKT, both in the PI3K signaling pathway, identifying genes consistently modulated by two different perturbations in the same pathway during tumorigenesis [hypergeometric P-value (HP) 10−27] (22,26). (B) Increasing overlap of castration signature with PTEN-knockout-driven prostate cancer over time (minimum HP 10−27) (22,27). Each map shows the overlap between non-cancerous prostate tissue following castration in a mouse (compared to prostate tissue from uncastrated mice) and a mouse model of prostate cancer driven by PTEN loss (compared to prostate tissue from wild-type littermates). The degree of overlap increases with time and is reversed when testosterone is given for 3 days. (C) Significant overlap between a stem cell signature from murine mammary glands and BCR-ABL fusion onco-protein signature (HP 10−8) (30,31). (D) High overlap between gene-expression changes driven by overexpression of KRAS2 oncogene in a cell line and by mutated KRAS2 in human lung tumors, identifying signaling events downstream of this oncogene that are potentially clinically relevant and can be studied in the cell line model (HP 10−24) (35). (E) High overlap between treatment with the small-molecule inhibitor imatinib in two leukemia patient populations, identifying higher confidence markers for drug response (HP 10−141) (21,36). (F) Significant overlap between cell line-derived profiles from different tissue types based on sensitivity to the small-molecule inhibitor dasatinib, identifying possible adaptation mechanisms which can be targeted for higher drug efficacy (HP 10−33) (38,39). The tick mark on each axis indicates the point in the ranked gene-expression signature list where the direction of differential expression switches. A permutation P-value based on the sum of the signal in bottom left and top right regions (perm pval) or when permutations are not possible the maximum of the BY-corrected RRHO map (max BY) is indicated.
Figure 3. RRHO identifies statistically significant overlap between expression signatures supporting or generating biological hypotheses. (A) High overlap in gene-expression changes between mouse models of prostate neoplasia driven by PTEN and AKT, both in the PI3K signaling pathway, identifying genes consistently modulated by two different perturbations in the same pathway during tumorigenesis [hypergeometric P-value (HP) 10−27] (22,26). (B) Increasing overlap of castration signature with PTEN-knockout-driven prostate cancer over time (minimum HP 10−27) (22,27). Each map shows the overlap between non-cancerous prostate tissue following castration in a mouse (compared to prostate tissue from uncastrated mice) and a mouse model of prostate cancer driven by PTEN loss (compared to prostate tissue from wild-type littermates). The degree of overlap increases with time and is reversed when testosterone is given for 3 days. (C) Significant overlap between a stem cell signature from murine mammary glands and BCR-ABL fusion onco-protein signature (HP 10−8) (30,31). (D) High overlap between gene-expression changes driven by overexpression of KRAS2 oncogene in a cell line and by mutated KRAS2 in human lung tumors, identifying signaling events downstream of this oncogene that are potentially clinically relevant and can be studied in the cell line model (HP 10−24) (35). (E) High overlap between treatment with the small-molecule inhibitor imatinib in two leukemia patient populations, identifying higher confidence markers for drug response (HP 10−141) (21,36). (F) Significant overlap between cell line-derived profiles from different tissue types based on sensitivity to the small-molecule inhibitor dasatinib, identifying possible adaptation mechanisms which can be targeted for higher drug efficacy (HP 10−33) (38,39). The tick mark on each axis indicates the point in the ranked gene-expression signature list where the direction of differential expression switches. A permutation P-value based on the sum of the signal in bottom left and top right regions (perm pval) or when permutations are not possible the maximum of the BY-corrected RRHO map (max BY) is indicated.CC BY-NC · PubMed Central
Figure 4. RRHO yields comparable significance results to GSEA while adding a 2D perspective. (A) Schematic showing how RRHO and GSEA have been applied for comparing two continuous expression signatures. A previous application of GSEA used statistical methods (SAM algorithm) to choose an appropriate fixed threshold to define two ‘gene sets’ of up- and downregulated genes from their mouse model of lung cancer. These gene sets were then used in GSEA to search for overlap with the expression profile from human tumors treated as a continuous lists (SAM/GSEA approach) (1). RRHO treats both signatures as continuous by stepping through all possible threshold pairs in both ranked gene lists to create a hypergeometric overlap map. The two algorithms use different but highly related enrichment statistics, the hypergeometric distribution P-value for RRHO and the Kolmogorov–Smirnov statistic for GSEA (see ‘Materials and Methods’ section). Both approaches employ multiple hypothesis corrections through either permutations or analytical corrections, performing the sliding threshold(s) step in all permutations. (B and C) RRHO maps showing the overlap between a KRAS2-driven mouse model profile (x-axes) and two human lung cancer profiles (y-axes), one for lung tumorigenesis in general and one specific to KRAS2 status within tumor subgroups. Red dashed lines approximate the threshold chosen to represent the mouse model gene set in the published GSEA-based analysis and relate to the mountain plot output produced by GSEA. (D) Comparison of permutation P-values between the SAM/GSEA and RRHO approaches. Permutation P-values from our best recreation of Sweet-Cordero SAM/GSEA analysis mouse model defined up and down (dn) gene sets are listed juxtaposed to permutation P-values from summary statistic-based interpretations of the RRHO maps for the same data. Up results are from analysis of genes upregulated in both experiments, and likewise for down. Up+dn RRHO results simultaneously consider both directions by calculating the sum of the maximal absolute log P-value from both the up–up and down–down regions of the heatmap during the permutation analysis (see ‘Materials and Methods’ section).
Figure 4. RRHO yields comparable significance results to GSEA while adding a 2D perspective. (A) Schematic showing how RRHO and GSEA have been applied for comparing two continuous expression signatures. A previous application of GSEA used statistical methods (SAM algorithm) to choose an appropriate fixed threshold to define two ‘gene sets’ of up- and downregulated genes from their mouse model of lung cancer. These gene sets were then used in GSEA to search for overlap with the expression profile from human tumors treated as a continuous lists (SAM/GSEA approach) (1). RRHO treats both signatures as continuous by stepping through all possible threshold pairs in both ranked gene lists to create a hypergeometric overlap map. The two algorithms use different but highly related enrichment statistics, the hypergeometric distribution P-value for RRHO and the Kolmogorov–Smirnov statistic for GSEA (see ‘Materials and Methods’ section). Both approaches employ multiple hypothesis corrections through either permutations or analytical corrections, performing the sliding threshold(s) step in all permutations. (B and C) RRHO maps showing the overlap between a KRAS2-driven mouse model profile (x-axes) and two human lung cancer profiles (y-axes), one for lung tumorigenesis in general and one specific to KRAS2 status within tumor subgroups. Red dashed lines approximate the threshold chosen to represent the mouse model gene set in the published GSEA-based analysis and relate to the mountain plot output produced by GSEA. (D) Comparison of permutation P-values between the SAM/GSEA and RRHO approaches. Permutation P-values from our best recreation of Sweet-Cordero SAM/GSEA analysis mouse model defined up and down (dn) gene sets are listed juxtaposed to permutation P-values from summary statistic-based interpretations of the RRHO maps for the same data. Up results are from analysis of genes upregulated in both experiments, and likewise for down. Up+dn RRHO results simultaneously consider both directions by calculating the sum of the maximal absolute log P-value from both the up–up and down–down regions of the heatmap during the permutation analysis (see ‘Materials and Methods’ section).CC BY-NC · PubMed Central
Figure 5. Using RRHO to survey a compendium of gene-expression signatures. (A) We used both RRHO and a GSEA-based approach to compare the overlap of a gedunin treatment gene-expression profile with a panel of small molecule drug treatment signatures from the Cmap database. Mean Cmap scores are reported for multiple instances of a single drug (x-axis) using a gedunin signature of 50 up- and 50 downregulated genes. Benjamini-Yekutieli (BY)-corrected RRHO maps for multiple instances were created individually and combined into a composite map using a pixel-by-pixel mean; the maximum positive or minimum negative value (whichever has higher absolute value) is reported (y-axis). The Spearman rank correlation coefficient between the overlap measures of these two scoring techniques shows overall positive correlation (ρ = 0.46). Estimated significance thresholds for each algorithm are indicated with red dashed lines. (B–E) Mean BY-corrected RRHO maps showing gene-expression overlap of gedunin with the HSP90 inhibitors celastrol (B) and 17-AAG (C), the lysosome uptake inhibitor monensin (D), and the PI3 kinase inhibitor wortmannin (E). These drugs show statistically significant overlap by RRHO analysis, but varying degrees of overlap by Cmap analysis. Note that in the two cases where RRHO analysis detects signal while the GSEA-based approach does not (D and E), the maximal overlap is outside the top 50 genes used in the GSEA-based approach (i.e. the signal is further from the lower left or upper right corner than in the other cases).
Figure 5. Using RRHO to survey a compendium of gene-expression signatures. (A) We used both RRHO and a GSEA-based approach to compare the overlap of a gedunin treatment gene-expression profile with a panel of small molecule drug treatment signatures from the Cmap database. Mean Cmap scores are reported for multiple instances of a single drug (x-axis) using a gedunin signature of 50 up- and 50 downregulated genes. Benjamini-Yekutieli (BY)-corrected RRHO maps for multiple instances were created individually and combined into a composite map using a pixel-by-pixel mean; the maximum positive or minimum negative value (whichever has higher absolute value) is reported (y-axis). The Spearman rank correlation coefficient between the overlap measures of these two scoring techniques shows overall positive correlation (ρ = 0.46). Estimated significance thresholds for each algorithm are indicated with red dashed lines. (B–E) Mean BY-corrected RRHO maps showing gene-expression overlap of gedunin with the HSP90 inhibitors celastrol (B) and 17-AAG (C), the lysosome uptake inhibitor monensin (D), and the PI3 kinase inhibitor wortmannin (E). These drugs show statistically significant overlap by RRHO analysis, but varying degrees of overlap by Cmap analysis. Note that in the two cases where RRHO analysis detects signal while the GSEA-based approach does not (D and E), the maximal overlap is outside the top 50 genes used in the GSEA-based approach (i.e. the signal is further from the lower left or upper right corner than in the other cases).CC BY-NC · PubMed Central

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

Show all 55 references

Cited by

See all 487

Metadata

Authors
Plaisier, S. B., Taschereau, R., Wong, J. A., & Graeber, T. G.
Affiliations
California NanoSystems Institute (US)
Corresponding author
Graeber, Thomas G.
Venue
Nucleic Acids Research
Published
2010-07-21
Volume
38
Issue
17
Pages
e169-e169
ISSN
  • 0305-1048 (ISSN-L)
  • 1362-4962
OpenAlex
W2111065008
Cited by
468
References
55
Citations by year
2026: 36 · 2025: 41 · 2024: 66 · 2023: 57 · 2022: 40 · 2021: 57 · 2020: 51 · 2019: 34 · 2018: 27 · 2017: 17 · 2016: 12 · 2015: 9 · 2014: 6 · 2013: 4 · 2012: 4
FWCI
1.95
Type
journal article
Language
en
Open access
Locations
  • Nucleic Acids Research (publishedVersion, cc-by-nc)
  • Europe PMC (PubMed Central) (publishedVersion, cc-by-nc)
  • PubMed Central (publishedVersion, cc-by-nc)
  • PubMed (publishedVersion)
  • eScholarship (California Digital Library) (submittedVersion, other-oa)
  • eScholarship (California Digital Library) (submittedVersion)
  • HighWire Press Open Archive (submittedVersion)
Funders
  • U.S. Department of Health and Human Services
  • Alfred P. Sloan Foundation
  • National Cancer Institute (T32CA09056, T32CA009056)
  • National Institute of General Medical Sciences (T32GM008652)
Topics
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Bioinformatics and Genomic Networks
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Gene expression and cancer classification
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Genomics and Phylogenetic Studies
Keywords
tumor progression, stem cell properties, web-based analysis tool
MeSH terms
Algorithms, Animals, Computer Graphics, Data Interpretation, Statistical, Humans, Neoplasms, Gene Expression Profiling, Databases, Genetic, Mice

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Rank–rank hypergeometric overlap: identification of statistically significant overlap between gene-expression signatures. https://app.cochat.ai/papers/10.1093/nar/gkq636
The paper
Plaisier, S. B., Taschereau, R., Wong, J. A., & Graeber, T. G. (2010). Rank–rank hypergeometric overlap: identification of statistically significant overlap between gene-expression signatures. Nucleic Acids Research. https://doi.org/10.1093/nar/gkq636

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.