ChiTaRS-3.1—the enhanced chimeric transcripts and RNA-seq database matched with protein–protein interactions

Updated database content

In the current 2016 update, 34 922 chimeric transcripts have been collated from eight organisms (Table 1). We have identified and annotated 11 714 cancer breakpoints from the recent study of Merten et al and from the Mitelman collection (39–42). To study all these cancer fusions (see ‘Breakpoints’ collection), we have performed manual confirmation of their veracity using the information from >7700 PubMed articles and >19 000 iHop links (Table 1 and Figure 1). Malignancies with the most frequently found fusions are Adenocarcinoma (6308 fusions, ChiTaRS-3.1), Chronic Leukemia (1140 fusions), Acute Lymphoblastic Leukemia (2078 fusions), and Acute Myeloid Leukemia (AML) (135 fusions) (Supplementary Table S2). ChiTaRS-3.1 consists of 435 chimeric transcripts and their junction sites that have been confirmed by RNA-seq datasets (the Human Body Map dataset analyses from (18)), and 77 chimeras have been confirmed by the mass-spec experiments (18,33,34). Finally, for all the Breakpoints collection, the website tool has been greatly improved to provide a user-friendly interface (see ‘Breakpoints’, and Supplementary Figure S1).

To make the ChiTaRS-3.1 collection the most comprehensive source of chimeric transcripts available, we regularly update the list of chimeras deriving from the GenBank collection of ESTs and mRNAs for H. sapiens (UCSC reference genome: GRCh37/hg19), M. musculus (NCBI37/mm9), D. melanogaster (BDGP R5/dm3), R. norvegicus (RGSC Rnor_6.0/rn6), B. taurus (Baylor College of Medicine HGSC Btau_4.6.1/bosTau7), D. rerio (Sanger Institute Zv9/danRer7), S. cerevisiae (SGD April 2011 sequence/sacCer3) and Sus scrofa (Broad/Pig3) (35,49). Over the past two years, we have added additional chimeric transcripts for all eight organisms: H. sapiens (22 810), M. musculus (7457), D. melanogaster (2740), R. norvegicus (10), B. taurus (6), D. rerio (7), S. cerevisiae (5), S. scrofa (89) (Table 1).

To provide biological context to fusion sequence data, we updated the GenBank (35), RefSeq (48) and Mitelman (39–42) cross-references for all the genes (total 170 797 cross-references). All the UniProt (43) references into ChiTaRS-3.1 have been updated and include now >17 000 unique proteins. Further, we added 29 643 Ensembl (46) cross-references for all gene names (Table 1). Finally, we added links to GeneCards (44), iHop (45), OMIM (47) and PubMed publications for 34 922 entries in the ‘Full Collection’ (Figure 1). Thus, ChiTaRS-3.1 is an easy-to-use resource for the in-depth study of fusion transcripts and proteins on a genome-wide, and multi-species level.

Updated database functionality

The improved user interface of ChiTaRS-3.1 allows for rapid and easy analysis of evolutionary conservation of fusions, literature references and experimental data supporting fusion expression in different organisms (see ‘Compare and Analyze’). We added a separate pop-up window with an extended annotation for every entry in the Full Collection (Figure 1, see a green button of the fusion, “EU216064"), allowing easy cross-reference to other databases (listed above).

Annotation of sense-antisense (SaS) chimeras

The phenomenon of Sense-antiSense (SaS) chimeric transcripts (34) is also covered by the ChiTaRS database in this latest version. While SaSs may result in chimeric protein translation, they also represent potential inhibitors of translation though dsRNA-mediated mechanisms (34) (Supplementary Figure S2). SaSs that have been identified in any of the eight organisms in ChiTaRS-2.1 (34) are easily accessed by clicking a check-box (‘Sense-ANTIsense transcripts’) on the ‘Full Collection’ page (Supplementary Figure S2). We collected 6485 chimeric RNA transcripts found in the eight organisms that comprise sense and antisense exons of the same open reading frame, incorporating them into ChiTaRS-3.1 (Table 1). Moreover, we have added manual annotation for 200 SaS chimeras that includes predicted trans-membrane domains of the translated fusion (in six frames), number of overlapping exons at the chimeric junction site, onco-genes, and the corresponding (homologous) transcripts found in other seven organisms. Interestingly, we have identified 17 common SaS chimeras in H. sapiens, M. musculus and D. melanogaster as well as 11 evolutionary-conserved SaS chimeras in three organisms: H. sapiens, M. musculus and S. scrofa (Supplementary Figure S3). Curiously, we found that two-thirds of the genes in those SaS chimeras are evolutionary-conserved phosphatases that lose their phosphate binding sites by means of incorporation of antisense exons (data not shown). This observation appears to be in line with our previous findings that some chimeric proteins tend to have a dominant-negative function in cells (19). Thus, ChiTaRS-3.1 uniquely and comprehensively catalogs SaS chimeras, rendering the study of their evolution and function readily accessible for users, world-wide.

Chimeric protein–protein interaction (ChiPPI) networks

Next, we asked how fusion protein function can be best assessed using computational methods. A fusion protein typically contains discreet domains from both parental proteins. It has been shown that a fusion event can dramatically alter the protein–protein interaction (PPI) network of the parental proteins. We, therefore, designed a visual means of assessing PPI network perturbations induced by protein fusion events. To this end, we have assembled a tool for analyzing PPI networks that focuses on individual protein domains as the mediators of PPIs (http://chippi.md.biu.ac.il/). Using this tool we have fully pre-computed chimeric protein–protein interactions (ChiPPI) networks for 2081 fusion proteins, and predicted ChiPPI networks for 22 527 human chimeric transcripts. The predicted ChiPPI networks have been produced by the unification of the PPI networks of two parental proteins of a chimera (see ‘Full Collection’). This new feather provides users with the ability to study the protein interaction networks of chimeric proteins for all cancer fusions.

ChiPPI displays PPI networks in a map that gives a broad overview of the consequences of a fusion event from a proteomic perspective. ChiPPI predicts where fusion proteins are likely to lose binding to interactors of the parental proteins. Figure 2 shows the predicted PPI network of AFF1-KMT2A fusion protein. The interactions of KMT2A with the tumor suppressors: MEN1, SMARCB1 and CBFB, as well as the interactions of AFF1 with the tumor suppressors: EAF1 and SIAH1 are lost upon the fusion. Using ChiPPI, we mapped the influence of specific fusion proteins on cellular PPI networks and on essential pathways in cancer development and progression. For example, Supplementary Table S3 shows the mapping of ChiPPI networks for the different ABL1 fusions and their alterations in the ‘betweenness centrality’, ‘clustering coefficient’, and scoring for the addition of onco-proteins to, or removal of tumor suppressors from, the PPI network. In general, we find that the PPI networks of fusion proteins often lose tumor suppressor proteins, as well as being enriched in onco-proteins. Thus, ChiPPI is highly suitable for displaying how fusion proteins contribute to the skewing of protein interaction networks as well as of signaling pathways. Particularly, this new feature provides users with the ability to study the protein interaction networks of different cancer fusions (http://chippi.md.biu.ac.il/).