LIACS Hallmark of Cancer Homepage

Supplemental Files of the research 'Mapping Cancer Hallmarks to Genes: Comparing the Impact of Different Approaches for Pan-Cancer Analyses'(In review)

In this study, we investigate the variation between published hallmark-gene mapping schemes and the impact this has on pan-cancer research results. We reused and re-analyse data from a recent pan-cancer study, changing the hallmarks mapping scheme and keeping all other analysis components the same. The result of our investigation shows large inconsistences in the way that hallmarks are defined. This leads to differences in definition of what constitutes a hallmark gene and in what the collective set of hallmark genes can show us about similarities between cancer types

Section I

In this section, we firstly showed the conceptual workflow of our research. Then we showed the results of comparison of Gene datasets generated by using different hallmark mapping methods in different aspect.

	Figure 1 present the conceptual workflow of our study, GO1~4 represent 4 Gene datasets generated by different hallmarks mapping methods. PW1 represents a gene datasets generated by a pathway mapping methods.Figure 1		Figure 2A the upset plot presents the intersections between hallmark datasets generat-ing by different mapping methods. Bar at the upper sides represent the intersections exclusively between selected datasets.bars colored in red represent gene shared by all of the datasets while bar colored in orange represents the number of genes exclusively shared by the 4 GO mapping methods Figure 2A
	Figure 2B Semantic Similarity between Gene Ontology Mapping Methods. Each box represents the score of pairwise comparison of two Gene Ontology lists. The depth of color shows the similarities between the two Gene Ontology set. Figure 2B		Figure 2C The Number of Member Genes Each Hallmark. GO1, GO2, GO3 and GO4 represent the 4 Gene ontology mapping methods. The height of bars represents the number of genes belonging to specific hallmarks. H1 to H10 represent hallmarks: 1) Sustaining Proliferative signaling, 2) Evading Growth Suppressor, 3) Resist Cell Death, 4) Enabling Replicative Immortality, 5) Inducing Angiogenesis,6) Activating Invasion& Metastasis, 7) Genome Instability & Mutation, 8) Tumor Promoting Inflammation, 9) Deregulating Cellular Energetics and 10) Avoiding Immune Destruction. Figure 2C
	The proportion of Go terms selected by Gene Ontology mapping methods. One, Two, three represent GO terms only selected by 1, 2, 3 GO mapping methods, respectively. The proportion represent one class of GO terms accounted for in specific hallmarks. The corresponding hallmarks of H1 to H10 can be found in Figure 2C Figure 2D

Section II

After knowing the impact of different mapping methods on Gene datasets composition, we next want to know whether these difference could influence the conclusion of researches that using Hallmark gene datasets in their analysis.We re-analysis part of Ulhen's research by applying the 4 GO mapping methods we mentioned before and see how the conclusion of research were affected

Figure 3A and 3B Prognostic-hallmark genes shared by multiple cancers. The Blue nodes represent prognostic-hallmark genes and the red nodes represent different cancer types. Edges linked Cancer and Genes means these genes were seen as prognostic-hallmark genes in these cancer type. For more information you can find in NDEx links Figure 3A Figure 3B

The co-expression genes network could be seen in Ndex using GO1 methods GO1
The co-expression genes network could be seen in Ndex using GO2 methods GO2
The co-expression genes network could be seen in Ndex using GO3 methods GO3
The co-expression genes network could be seen in Ndex using GO4 methods GO4
The co-expression genes network could be seen in Ndex using PW1 methods PW1

Figure 4A and 4B show cluster enrichment Analysis of Lung cancer for the GO1 and GO4 hallmark dataset. Red nodes represent enrichment in both hallmark and prognostic genes, orange represents prognostic-only enrichment and grey represents hallmark-only enrichment. The edges represent the correlations between each cluster Figure 4A&4B

Figure 4C,4D,4E and 4F show the proportion of hallmarks that prognostic-hallmark clusters are enriched in. The corresponding hallmarks of H1 to H10 can be found in Figure 2C Figure 4C Figure 4D Figure 4E Figure 4F

Figure 4G and 4H The two graphs show the proportion of different groups of prognostic genes when applying GO1 and GO4. Red bar represents genes co-expressed hallmark genes. Yellow bar represents prognostic-hallmark genes. Blue bar represents prognostic genes neither co-expressed with nor defined as hallmark genes marks may provide insight. Figure 4G Figure 4H

Section III

After investigating the impact on research conclusion by applying multiple mapping methods to re-analysis an study from Ulhen's research, we already noticed that due to the lack of shared-understanding on what hallmarks is,self-defined mapping methods could lead to different results. As we mentioned before, 793 genes were defined as hallmark genes in all methods.his consensus may represent a core set of genes that can be linked to cancer hallmarks with a greater confidence.

Figure 5 The enriched analysis of core gene sets. The two-layer full hierarchical struc-ture consists of an inner ring, which represents parent GO terms and the outer ring, which represents child GO termsFigure 5