Structural Variants (SV) Manta output interpretation: MantaBND and translocation lenght

Question

[nihilior]

Hi everyone,

Manta novice here. Does anybody know how is possible to retrieve the translocation lenght from Manta SV output (if it makes sense)? For example, we have the following translocation:

chr1 - chr6 : chr1 95016699 ]chr6:158767194]A MantaBND:4649:0:1:0:0:0:1

chr6 - chr1: chr6 158767194 C[chr1:95016699[ MantaBND:4649:0:1:0:0:0:0

and we would need to know the lenght of the translocated fragment. Also, can someone examplain us how the "alt" column is structured? Like, how must we interpret the following different cases:

ATTACAGGCGTGAGCCACCGCGCCC]chr6:78364271] ;
]chr6:158767194]A ; C[chr1:95016699[

Thanks so much in advance

Answer 1

[cmdcolin]

These two entries that you list

chr1 95,016,699 ]chr6:158,767,194]A MantaBND:4649:0:1:0:0:0:1
chr6 158,767,194 C[chr1:95,016,699[ MantaBND:4649:0:1:0:0:0:0

They are reciprocal. It is maybe not meaningful to know any length attribute from just these two entries

I use this ascii art to help me understand breakends, a feature on chr13 at position 123 has a breakend C[2:456[ then it looks like this schematically: after a C on chr13, it connects to chr2 at position 456 and "keeps going" to the right of that breakpoint on chr2 (It could also go to the left of that point if the breakend syntax used square brackets pointing to the left e.g. ] see https://samtools.github.io/hts-specs/VCFv4.3.pdf for more detail)

    chr13:123
  -------------C\
                 \
                  \
                   \
                    \
                     \
                      \
                       \
                        \--------------
                         chr2:456

If you want to try to understand your breakends in more detail, you can consider trying jbrowse 2 (https://jbrowse.org). We have a tool called breakpoint split view that draws the read evidence and VCF evidence for your breakends https://jbrowse.org/jb2/docs/user_guide/#launching-breakpoint-split-view as well as a circular whole genome overview https://jbrowse.org/jb2/docs/user_guide/#example-sv-inspector-workflow (I'm a developer on jbrowse 2)

[nihilior]

Thank you very much for the reply cmdcolin, it was very helpful. I will absolutely try jbrowse 2 and the breakpoint split view, it was kinda exactly what we needed. But it is still hard to me to understand why the lenght of my translocation is not meaningful. Maybe I will understand better using your suggested tools. Thanks again

[cmdcolin]

If you can formulate what length represents, it could possible be pursued, but without a definition of the length, it's hard to say. For example if it is a balanced translocation (https://genetics.thetech.org/sites/default/files/BT1ll.jpg) possibly the chromosome arms could be exchanged (and you could say the "length" is related to the length of that chromosome arm). But, on a bioinformatic level, the VCF breakends are sort of hard to interpret as there may be other breakend events so we cant just say that there is a chromosome arm exchanged by looking at those two events (for example if there is another breakend further along in the chromosome, it might suggest there is something else going on, e.g. just a smaller chunk of chr6 is placed into chr1 or something like this).

Answer 2

[d-cameron]

we would need to know the lenght of the translocated fragment.

A 'translocated fragment' requires at least two breakpoints and you have only listed one. If this is truely an isolated breakpoint then there is no meaningful 'length' - you'll have a derivate chromosome that goes from the telomere of chromosome 1, joins to chromosome 6 and goes all the way to the telomere of chromosome 6.

Also, can someone examplain us how the "alt" column is structured?

This is defined in Section 5.4 of the VCF file format specifications: https://samtools.github.io/hts-specs/VCFv4.3.pdf I strongly recommend reading all of section 5.

A critical aspect of SV interpretation frequently missed by those new to the field is that SV callers report breakpoints. Compex genomic rearrangements are made up of multiple breakpoints but existing SV callers (such as manta) onlydo breakpoint detection. They do not do genomic rearrangement event interpretation. This is a separate step downstream of SV calling and there are very few tools that do this. If you are interested in understanding genomic rearrangements, I recommend reading the LINX preprint (https://www.biorxiv.org/content/10.1101/2020.12.03.410860v1). Even if you just skip to the end to look at the figures, it should quite quickly become clear that you can't take a single translocation in isolation - you need to look at the full set of breakpoints and copy number changes to be able to reconstruct the derivate chromosomes and interpret complex events. SV callers such as Manta don't do this- they only detect the breakpoints.

[nihilior]

Thanks so much d-cameron, your answer was very helpful to help me understand what's going on. I did not fully understand what was the real purpose of the SV caller. I will read the preprint and try to use LINX. Thanks again

[Alessandro]

Ok , thank you so much !

Answer 3

[Alessandro]

Hi everybody, I have the same question, in particular I would like to verify some translocations observed with FISH exp. In particular, I need to understand:

• Why does the VCF annotation not show the length for this type of SVs?
• how to determinate the length of translocation (BND);
• The different meaning in the ALT annotation, for example: why in some cases the nucleotide is written first: C[chr1:95,016,699[ and in other cases the nucleotide is written at the end of the string : ]chr6:158,767,194]A;

[d-cameron]

The different meaning in the ALT annotation

These encode the orientation of the breakpoint. See Section 5.4 of the VCF file format specifications: https://samtools.github.io/hts-specs/VCFv4.3.pdf I strongly recommend reading all of section 5.

Why does the VCF annotation not show the length for this type of SVs?

Because an isolated interchromosomal breakpoint doesn't have a meaningful length.

how to determinate the length of translocation (BND)

You can't because there's no length. A BND record is not a 'translocation', it is an assertion that a breakpoint exists. That is, the two pieces of DNA adjacent to the breakpoint positions are directly connected to each other in the sample. That is all a BND call is. The SV callers GRIDSS/GRIDSS2 and HYDRA report every event in BND notation. A DEL call from Manta is identical to an intrachromosomal +/- oriented BND call in GRIDSS. They've both claiming that a breakpoint in an orientation consistent with a deletion exists. Note that neither Manta nor GRIDSS are claiming that the DNA in the 'deleted' segment is actually deleted, merely that the DNA before&after the 'deleted' segment are adjacent. VCF version 4.4 adds an SVCLAIM field to disambiguate between a breakpoint DEL call, a copy number DEL call, and an actual deletion.

A translocation in the sense of a piece of DNA of a given length from one chromosome being inserted into another involves two breakpoints. SV callers only report the presence of breakpoints. Actually determining that a translocation of a given size exists requires interpretation of the set of breakpoints (and CNV calls) and is not done by SV callers. It's done by genomic rearrangements classification/interpretation tools such as LINX (https://www.biorxiv.org/content/10.1101/2020.12.03.410860v1.full.pdf). As you can see from the figures at the end of that preprint, genome rearrangements can get very complicated - especially in cancer.

[cmdcolin]

You may have to come up with what you think is a biological meaning for "length of SV", a breakend may tells you that chr1 is joined to chr6 at a particular point, but you may need further downstream analysis to determine if that means there is a balanced translocation of entire chromosome arms or something of that nature

[Alessandro]

Thank you for your answer. Since BND are translocations, in the example :

chr1 95,016,699 ]chr6:158,767,194]A

chr6 158,767,194 C[chr1:95,016,699[

I expect to find a fragment of chromosome 1 on chromosome 6. In this case it should be possible to calculate the length of this translocated fragment. This is my point.