Understanding cancer neochromosome formation through integrated analysis of fusion and copy number data

Presented by: Dr Vincent Corbin


Neochromosomes (NCs) are “extra” chromosomes that are found in around 3% of tumour genomes. They are a hallmark of liposarcoma, which is a cancer of the fat cells studied by the Papenfuss lab at WEHI in conjunction with colleagues at the Peter MacCallum Cancer Centre. Circular chromosomes and giant rod chromosomes are both examples of NCs and are comprised of multiple donor segments from other regions of the region that are frequently highly amplified.

Until now, NCs have only been studied with low resolution technologies such as Fluorescence In Situ Hybridisation (FISH) and microarrays. The Papenfuss lab have been studying a cell-line derived from a patient with liposarcoma. The DNA was flow-sorted to enrich for the NCs (which is larger than the regular chromosomes) and then sequenced using the Illumina HiSeq and GAIIx.


They generated 300 million reads across multiple sequencing runs that were aligned against the human reference genome. Dr Arthur Hsu, a post-doc in the Papenfuss lab, has been studying the sequence to estimate breakpoints in the neochromosome (breakpoints occur where two derived contiguous sequences are joined in the neochromosome) and Vincent has been using the data to estimate copy number variation (CNV).

The data reveals that the amplified regions of the NC contain known drivers of liposarcoma tumours. Most chromosomes contribute material to the NC and most of these regions are highly amplified (16x-32x), with the exception of regions comprising the telomere and centromere of the NC. This appears as small, highly amplified islands in plots of copy number for each contributing chromosome. There is strong evidence that chromosome 12 was the origin of the NC and that extra material was acquired later in development. 

The breakpoints can be classified as edge-to-edge, edge-to-interval and interval-to-interval. Interestingly, the copy number ratio does not vary across edge-to-edge boundaries. All of this suggests a breakage-fusion-bridge (BFB) model may be appropriate for the development of the NC or, perhaps more plausibly, chromothripsis (chromosome shattering, followed by the cell trying to rejoin the pieces but inevitably in the wrong order). 


Vincent is working on a mathematical model for the BFB model and simulating observations from the model to compare against the data. He presented three models for the breakpoint process:

  1. Uniform
  2. Small duplications/deletions
  3. Large duplications/deletions.

Vincent’s simulation results from the small duplications/deletions model closely resembled the data. There were four main criteria he used to compare the models against the data.

  1. This model generated small, highly amplified islands as seen in the data.
  2. The density plot of CN distribution in the data reveals the distribution is multi-modal with the two strongest peaks at 16x and 32x. However, none of the density plots from any of the three breakpoint models resemble the data.
  3. The 1:1 CN ratio across edge-to-edge breakpoints is captured by all three models, but Vincent suggested this is an artefact of the simulation process rather than evidence of a good model.
  4. A Circos plot of the inter-chromosomal re-arrangements. The uniform model performed better than the small duplications/deletions model.


In summary, all features of the data can be captured by the models but no one model can capture all features simultaneously. The large duplications/deletions model is limited by the size of the NC it creates. Vincent is exploring a size penalty to force this model to choose smaller ring chromosomes. This penalised model captures (1), (2) and (4), however (2) appears and disappears over time (i.e. the peak-modes moves) suggesting it may be a transient effect even in the data.

The lesson learnt from the (ongoing) simulation study is that chromothripsis may not be required and BFB may be sufficient to explain the development of the NC. But the orientation of the islands changes and the circular BFB can’t account for different orientations of the islands. Vincent posited the following explanations:

  • Linear BFB may be involved in addition to circular BFB
  • The circular BFB is poorly understood and thus may still admit the possibility of islands with different orientations.
  • Perhaps there was an early chromothripsis event follow by BFB. However the motifs around the breakpoints do not support this hypothesis.

The simulation model is ongoing work with current plans to extend it to include:

  • Simulate acquisition of new material by the NC
  • Exploring the linear BFB model
  • A population based model following multiple cells across time (the current simulation follows a single cell).