When the user inspects the results of fusion calls, there are 2 key aspects to verify:
- 1) The mappings towards the wild type genome
In order to inspect this, open the
Genome Browser View (WT)and add to the track list the original unaligned ends that are used to identify the breakpoints during the first round of the tool running: this track indicates the initial evidence that is used to process and refine the calls afterwards.
In the example shown in figure 4.5, we can see that a significant amount of evidence exists for unaligned ends to map on one side of the candidate fusion, but at the same time, if we inspect the other side of the fusion, we can see that no unaligned ends map in this locus. Additionally, a substantial amount of reads mapping on the left hand side of the fusion is ambiguously mapped (yellow).
The reason of the lack of reads on the right hand side of the fusion, can be followed up by inspecting the fusion chromosomes. If you wish to inspect the original mapping to the wild type, the original mapping in the first iteration folder can be added to this view: the reads that are later mapped to the fusion chromosome will appear here in their original location.
- 2) The corresponding mappings towards the fusion genome
Open the "Genome Browser View (Fusions)" (see figure 4.6). Here we can notice that many reads are mapping to the artificial chromosome (instead of mapping to the wild type one, where they were missing in the above view). However, the reads also in this case are not uniquely mapped.
This indicates one of those potential cases where a high degree of homology might exist in the genome between genes targeted by the panel (or for which the panel covers at least one side of the fusion): in this case clearly the reads are attracted by the artificial fusion chromosome, but due to the homology the lead to identify a fusion where instead there is not sufficient support. This is a clear example of false positive, due to high homology in the targets.
Note about known false positive fusions found when running QIAseq Targeted Fusion Panels
The following fusions can be disregarded as common read-through mRNAs or false fusions due to gene homology.
- HALC1-COLQ, common read through
- BCR-BCRP3 (BCRP3-BCR), known false positive
- TMP3-TMP4 (TMP4-TMP3), homologous genes
The fusions listed above are found when running QIAseq Targeted Fusion Panels; additional false positive fusions may be found when running custom made panels.
- 3) Promiscuity threshold
This value indicates the minimum number of different fusion partners that will define a gene as promiscuous (number of positions within the gene suggested by mapping of unaligned ends as in figure 4.7). The ready to use workflow has been designed to increase the sensitivity and therefore the promiscuity threshold has been set by default to 20 (but note that when using the tool on its own, the default is set to 7). If you expect samples with genes characterised by multiple fusion breakpoints, remember to increase the Promiscuity threshold value in order to capture also real fusions with multiple breakpoints within the same gene.