Dear organizers, Looking at the methods to generate the benchmarking datasets, it seems that the focus of the fusion challenge is on those chromosomal aberrations that fuse genes in proper orientation (5'-3') and that can hence yield chimeric proteins. However, there are many chromosomal aberrations that lead to aberrant transcripts which are not translated but which can still be relevant to tumor biology. For example translocations which truncate tumor suppressor genes lead to inactivation of these genes. Another example are translocations disrupting the 3' UTR of PD-L1, which has been shown to deregulate PD-L1 expression. Such chromosomal aberrations also manifest in the RNA: The former example yields transcripts with intergenic breakpoints; the latter example yields transcripts with breakpoints in the UTR. Some fusion detection tools also report these events in the results, because they are relevant to cancer. If only "classical" fusions are considered as true positives, this would put these tools at a disadvantage, because these events will inflate the result set and thus hurt precision. For the simulated and spike-in datasets, this is not an issue, since it is clear in these cases that no such events are expected. Those tools that usually report such events should simply disable this feature for these datasets. But on the challenge overview it says that in the end the tools will also be tested on real tumor data. How will these types of events be scored there? Other examples for aberrant transcripts which are not fusions in the classical sense but are nevertheless interesting in the context of cancer include: - chromosomal aberrations which fuse genes in nonsensical orientations (5'-5' or 3'-3'), such that the resulting transcript consists of the sense strand of one gene and the antisense strand of the other gene - small duplications inside a gene, yielding duplicate exons - small inversions inside a gene, leading to transcripts involving both the sense and antisense strand of a gene - small deletions inside a gene, leading to exon skipping On the other hand there are non-canonically spliced transcripts that look like they arise from chromosomal aberrations, but that are perfectly normal and are frequently observed in healthy tissue. Examples for such transcripts are: - cis-spliced transcripts between neighboring genes (a.k.a. read-through transcripts) - trans-spliced transcripts between genes on opposite strands - circular RNAs which look like duplications in NGS data - exon shuffling which leads to non-canonical ordering of exons and also looks like duplications in NGS data All of these can be observed in healthy tissue samples and are therefore unlikely to be of importance for cancer development. For this reason, some fusion detection tools remove such events from their results, even though a PCR-validation would probably turn out positive. Would this be rated as a false negative by the scoring system? This question is theoretically relevant to the simulated datasets, because FUSIM could - by chance - mimick a fusion between neighboring genes. Many thanks in advance for your clarification, Sebastian