(a) Enrichment analysis with a compendium of molecular concepts generates two sets of minimally overlapping signature concepts for fusion and point mutation genes. Molecular concepts are depicted as nodes with the size of each node corresponding to the number of genes in each concept. The thickness of the lines correlates with the significance of overlap tested by hypergeometric statistics. (b) The ConSig algorithm. Circles represent concepts associated with gene X. k is the total number of concepts associated with gene X, xi = number of fusion genes in concept i, ni = total genes in concept i. A corresponding figure for mutation genes is not shown. (c) Fusion and mutation ConSig scores for known fusion genes (red dots), cancer point mutation genes (blue dots) and all other human genes (gray dots). Genes known to be both fusion and mutation genes are purple. r, r.ConSig score; d, d.ConSig score, D-line, distinction line. (d) Identifying the top 60 genes rated by r.ConSig score (red column chart) produced a list highly enriched for established cancer genes (known fusion genes labeled with red stars; mutation genes blue stars). The d.ConSig scores are depicted by a blue line chart, with positive and negative values indicating the dots above or below the D-line. Known mutation or fusion genes matching the prediction by d.ConSig scores are marked by black dots in the line chart.

(a) Pair-end transcriptome sequencing of 12 lung cancer cell lines, followed by prioritizing the 3′ partners of paired-end chimeras (≥3 paired reads) by rConSig score (red bars). Left, EML4-ALK is a candidate fusion, and is supported by six paired reads (black bars), in the H2228 lung cancer cell line (known to harbor this fusion). Right, ConSig analysis nominates the R3HDM2-NFE2 fusion as the top candidate in the H1792 lung cancer cell line. (b) The R3HDM2-NFE2 fusion was confirmed by RT-PCR and sequencing of the PCR product. qRT-PCR of wild-type NFE2 revealed overexpression of NFE2 in a subset of lung cancer cell lines; only H1792 cells express the chimeric R2HDM2-NFE2. (c) Top, schematic of the genomic organization of R3HDM2-NFE2 fusion, with red and green bars indicating the location of BAC clones. This fusion was generated by an intrachromosomal translocation. Bottom, interphase FISH analysis showing amplification signals of 3′ NFE2 and 5′ R3HDM2 (left, middle) and R3HDM2-NFE2 fusion (right) on H1792 cell line. Normal signals are indicated by white arrows; aberrant colocalizing signals by yellow arrows; aberrant split signals by green or red signals. (d) Schematic of the R3HDM2-NFE2 fusion mRNA and protein. Structures for the R3HDM2 and NFE2 genes are derived from GenBank reference sequences. The numbers above the exons indicate the last base of each exon. Open reading frames are shown in darker shades. The exons of R3HDM2-NFE2 fusion are numbered from the original reference sequence. The lower schematic shows wild-type NFE2 protein and its domain architecture. (e) siRNA knockdown of NFE2 in H1792 cells leads to decreased cell proliferation (middle graph) and invasion (right graph). Percent knockdown of the fusion transcript revealed by qRT-PCR is shown in the left graph. (f) Analysis of SNP array data from 139 lung adenocarcinoma tissues revealed recurrent copy number aberrations in two patients at the 3′ NFE2 locus, as well as the focal amplification of R3HDM2-NFE2 fusion on H1792. (g) As in c, except the data are from three lung adenocarcinoma patients. L41 is a negative case with two colocalizing signals; L83 has split and high copy number gain at NFE2 locus; L18 showed one additional 3′ NFE2 signal (red).

(a) The hypergeometric statistics for assessing whether a group of fusion gene partners (e.g., all BCR partners) contains an unexpected number of genes that physically interact with the same gene (e.g., all genes that interact with PIK3R1). (b) The total number of significant links (589) and the number of fusion partner groups having these links (33) were plotted with the distribution calculated from randomly chosen gene sets with an equal amount of connectivity (1,000 permutations). (c) Analysis of the fusion partner groups with a compendium of molecular concepts by hypergeometric statistics. The numbers in the pie chart represent the number of significant concepts in each functional category (P ≤ 0.01). (d) Network visualization of the most significant (P < 10⁻⁷) instances where many fusion partners also interact with a shared gene. Fusion genes are green nodes; shared interacting genes are red (with color intensity indicating P-value). Red arrows designate gene fusions (from 5’ partners to 3’ partners); green lines represent molecular interactions. For simplicity, genes and proteins are both given in roman type in the diagram. For each fusion partner set, the shared interacting gene having the most significant P-value was designated as a fusion-interaction hub. Clusters are limited to known fusions joined by established molecular interactions. (i) Acute lymphoblastic lymphoma (ALL) fusions with a hub of GATA3. (ii) Acute/chronic myelogenous leukemia (AML and CML) fusions with a hub of MEIS1. (iii) B-cell lymphoma and chronic lymphoblastic lymphoma (CLL) fusions through the hubs of CDK6 and CTNNB1. (iv) AML fusions partially focusing on HDAC1; RUNX1 is the hub of immunoglobulin fusions, and also involved in multiple fusions in AML, and thus links clusters iii and iv. (v) ALL and CML fusions through the hub of PIK3R1. (vi) Sarcoma and prostate cancer fusions around ERG and MSK1. The PI3K3R1 and HDAC1 hubs are within dashed lines.

An SNP array data set was used to evaluate genomic aberrations associated with gene fusions in acute lymphocytic leukemia (ALL). (a) The recurrent TCF3-PBX1 fusion (n = 17) was associated with deletion of the 3’ region of TCF3 and duplication of the 3’ region of PBX1. Color scales indicate log2 transformed relative copy number. (b) Of the 56 samples with unbalanced gene fusions in this data set, 55 samples conformed to the fusion breakpoint principle. Fusion genes on their corresponding chromosomes were aligned with the unbalanced breakpoints. “U/A” represents the number of samples with unbalanced fusions out of the total samples with gene fusions. Colored bars in the 5’ partner and 3’ partner columns indicate copy number increases (red) or deletions (blue), and given in white is the number of samples having the DNA aberration. For example, the first row shows that out of 66 cases of ALL patients with BCR-ABL1 fusions, 14 cases are unbalanced, of which 6 have 5’ BCR duplication and 4 have 3’ BCR deletion. ABL1 can be interpreted in a similar fashion. (c) Unbalanced fusions in the 12 leukemia cancer cell lines follow the fusion breakpoint principle. The exceptions to the principle are marked by asterisks.