Limits...
Regulators of genetic risk of breast cancer identified by integrative network analysis.

Castro MA, de Santiago I, Campbell TM, Vaughn C, Hickey TE, Ross E, Tilley WD, Markowetz F, Ponder BA, Meyer KB - Nat. Genet. (2015)

Bottom Line: To better understand how risk loci might combine, we examined whether risk-associated genes share regulatory mechanisms.We identified 36 overlapping regulons that were enriched for risk loci and formed a distinct cluster within the network, suggesting shared biology.The risk transcription factors driving these regulons are frequently mutated in cancer and lie in two opposing subgroups, which relate to estrogen receptor (ER)(+) luminal A or luminal B and ER(-) basal-like cancers and to different luminal epithelial cell populations in the adult mammary gland.

View Article: PubMed Central - PubMed

Affiliation: Bioinformatics and Systems Biology Laboratory, Federal University of Paraná (UFPR), Polytechnic Center, Curitiba, Brazil.

ABSTRACT
Genetic risk for breast cancer is conferred by a combination of multiple variants of small effect. To better understand how risk loci might combine, we examined whether risk-associated genes share regulatory mechanisms. We created a breast cancer gene regulatory network comprising transcription factors and groups of putative target genes (regulons) and asked whether specific regulons are enriched for genes associated with risk loci via expression quantitative trait loci (eQTLs). We identified 36 overlapping regulons that were enriched for risk loci and formed a distinct cluster within the network, suggesting shared biology. The risk transcription factors driving these regulons are frequently mutated in cancer and lie in two opposing subgroups, which relate to estrogen receptor (ER)(+) luminal A or luminal B and ER(-) basal-like cancers and to different luminal epithelial cell populations in the adult mammary gland. Our network approach provides a foundation for determining the regulatory circuits governing breast cancer, to identify targets for intervention, and is transferable to other disease settings.

Show MeSH

Related in: MedlinePlus

Correlation of expression of targets shared between TF pairs in breast tumours(a-c) Correlations of gene expression between a given TF and its targets were plotted for three different TF-TF pairs as indicated. Above each panel a cartoon depicts the observed interactions. Red circles indicate co-activation, blue circles co-repression. Targets are shown in grey if the two TFs have opposing effects on the target. (d) Heat map of the correlation of gene expression for targets shared by any pair of the 555 TFs (cohort I, METABRIC) whose regulons were of sufficient size to be analysed in the EVSE pipeline. Unsupervised clustering was applied to this correlation heat map resulting in the dendrogram shown at the top of the plot. The black bars depict the 36 risk-TFs, which fall into two distinct clusters. (e) Enlargement of the correlation heat map for the risk-TFs only. Above the matrix a bar with yellow to red colouring depicts the results (BH adjusted p-values) of a MRA analysis for the enrichment within each regulon of positive and negative targets that are upregulated in ER+ or ER− tumours, respectively, in cohort I of the METABRIC samples. The panel to the left of the matrix shows the master regulators identified for the FGFR2 and E2 responses. (f) Relative gene expression levels of the risk-TFs in ER+ or ER− tumours in cohort I of the METABRIC samples: expression levels were averaged in all ER+ and all ER− tumours and compared to expression levels averaged across all samples. TFs are shown ranked by differential gene expression between ER+ and ER− tumours.
© Copyright Policy
Related In: Results  -  Collection


getmorefigures.php?uid=PMC4697365&req=5

Figure 4: Correlation of expression of targets shared between TF pairs in breast tumours(a-c) Correlations of gene expression between a given TF and its targets were plotted for three different TF-TF pairs as indicated. Above each panel a cartoon depicts the observed interactions. Red circles indicate co-activation, blue circles co-repression. Targets are shown in grey if the two TFs have opposing effects on the target. (d) Heat map of the correlation of gene expression for targets shared by any pair of the 555 TFs (cohort I, METABRIC) whose regulons were of sufficient size to be analysed in the EVSE pipeline. Unsupervised clustering was applied to this correlation heat map resulting in the dendrogram shown at the top of the plot. The black bars depict the 36 risk-TFs, which fall into two distinct clusters. (e) Enlargement of the correlation heat map for the risk-TFs only. Above the matrix a bar with yellow to red colouring depicts the results (BH adjusted p-values) of a MRA analysis for the enrichment within each regulon of positive and negative targets that are upregulated in ER+ or ER− tumours, respectively, in cohort I of the METABRIC samples. The panel to the left of the matrix shows the master regulators identified for the FGFR2 and E2 responses. (f) Relative gene expression levels of the risk-TFs in ER+ or ER− tumours in cohort I of the METABRIC samples: expression levels were averaged in all ER+ and all ER− tumours and compared to expression levels averaged across all samples. TFs are shown ranked by differential gene expression between ER+ and ER− tumours.

Mentions: To refine the clustering analysis and look for clues to biological function, we extended the RTN16 package (methods) to include the direction of association between any TF-target gene pair using Pearson correlation. For all pairs of TFs with a target gene in common, the correlation values were used to assess whether the TFs regulated shared target gene in the same direction (up or down), or in different (opposite) directions (Fig. 4a-c). This analysis was carried out for all TFs in our regulatory network, and the correlation heat map was used in unsupervised clustering to generate the dendrogram depicted above the matrix (Fig. 4d). The position of the 36 risk-TFs is highlighted by the black bars below the dendrogram.


Regulators of genetic risk of breast cancer identified by integrative network analysis.

Castro MA, de Santiago I, Campbell TM, Vaughn C, Hickey TE, Ross E, Tilley WD, Markowetz F, Ponder BA, Meyer KB - Nat. Genet. (2015)

Correlation of expression of targets shared between TF pairs in breast tumours(a-c) Correlations of gene expression between a given TF and its targets were plotted for three different TF-TF pairs as indicated. Above each panel a cartoon depicts the observed interactions. Red circles indicate co-activation, blue circles co-repression. Targets are shown in grey if the two TFs have opposing effects on the target. (d) Heat map of the correlation of gene expression for targets shared by any pair of the 555 TFs (cohort I, METABRIC) whose regulons were of sufficient size to be analysed in the EVSE pipeline. Unsupervised clustering was applied to this correlation heat map resulting in the dendrogram shown at the top of the plot. The black bars depict the 36 risk-TFs, which fall into two distinct clusters. (e) Enlargement of the correlation heat map for the risk-TFs only. Above the matrix a bar with yellow to red colouring depicts the results (BH adjusted p-values) of a MRA analysis for the enrichment within each regulon of positive and negative targets that are upregulated in ER+ or ER− tumours, respectively, in cohort I of the METABRIC samples. The panel to the left of the matrix shows the master regulators identified for the FGFR2 and E2 responses. (f) Relative gene expression levels of the risk-TFs in ER+ or ER− tumours in cohort I of the METABRIC samples: expression levels were averaged in all ER+ and all ER− tumours and compared to expression levels averaged across all samples. TFs are shown ranked by differential gene expression between ER+ and ER− tumours.
© Copyright Policy
Related In: Results  -  Collection

Show All Figures
getmorefigures.php?uid=PMC4697365&req=5

Figure 4: Correlation of expression of targets shared between TF pairs in breast tumours(a-c) Correlations of gene expression between a given TF and its targets were plotted for three different TF-TF pairs as indicated. Above each panel a cartoon depicts the observed interactions. Red circles indicate co-activation, blue circles co-repression. Targets are shown in grey if the two TFs have opposing effects on the target. (d) Heat map of the correlation of gene expression for targets shared by any pair of the 555 TFs (cohort I, METABRIC) whose regulons were of sufficient size to be analysed in the EVSE pipeline. Unsupervised clustering was applied to this correlation heat map resulting in the dendrogram shown at the top of the plot. The black bars depict the 36 risk-TFs, which fall into two distinct clusters. (e) Enlargement of the correlation heat map for the risk-TFs only. Above the matrix a bar with yellow to red colouring depicts the results (BH adjusted p-values) of a MRA analysis for the enrichment within each regulon of positive and negative targets that are upregulated in ER+ or ER− tumours, respectively, in cohort I of the METABRIC samples. The panel to the left of the matrix shows the master regulators identified for the FGFR2 and E2 responses. (f) Relative gene expression levels of the risk-TFs in ER+ or ER− tumours in cohort I of the METABRIC samples: expression levels were averaged in all ER+ and all ER− tumours and compared to expression levels averaged across all samples. TFs are shown ranked by differential gene expression between ER+ and ER− tumours.
Mentions: To refine the clustering analysis and look for clues to biological function, we extended the RTN16 package (methods) to include the direction of association between any TF-target gene pair using Pearson correlation. For all pairs of TFs with a target gene in common, the correlation values were used to assess whether the TFs regulated shared target gene in the same direction (up or down), or in different (opposite) directions (Fig. 4a-c). This analysis was carried out for all TFs in our regulatory network, and the correlation heat map was used in unsupervised clustering to generate the dendrogram depicted above the matrix (Fig. 4d). The position of the 36 risk-TFs is highlighted by the black bars below the dendrogram.

Bottom Line: To better understand how risk loci might combine, we examined whether risk-associated genes share regulatory mechanisms.We identified 36 overlapping regulons that were enriched for risk loci and formed a distinct cluster within the network, suggesting shared biology.The risk transcription factors driving these regulons are frequently mutated in cancer and lie in two opposing subgroups, which relate to estrogen receptor (ER)(+) luminal A or luminal B and ER(-) basal-like cancers and to different luminal epithelial cell populations in the adult mammary gland.

View Article: PubMed Central - PubMed

Affiliation: Bioinformatics and Systems Biology Laboratory, Federal University of Paraná (UFPR), Polytechnic Center, Curitiba, Brazil.

ABSTRACT
Genetic risk for breast cancer is conferred by a combination of multiple variants of small effect. To better understand how risk loci might combine, we examined whether risk-associated genes share regulatory mechanisms. We created a breast cancer gene regulatory network comprising transcription factors and groups of putative target genes (regulons) and asked whether specific regulons are enriched for genes associated with risk loci via expression quantitative trait loci (eQTLs). We identified 36 overlapping regulons that were enriched for risk loci and formed a distinct cluster within the network, suggesting shared biology. The risk transcription factors driving these regulons are frequently mutated in cancer and lie in two opposing subgroups, which relate to estrogen receptor (ER)(+) luminal A or luminal B and ER(-) basal-like cancers and to different luminal epithelial cell populations in the adult mammary gland. Our network approach provides a foundation for determining the regulatory circuits governing breast cancer, to identify targets for intervention, and is transferable to other disease settings.

Show MeSH
Related in: MedlinePlus