Skip to content

Latest commit

 

History

History
129 lines (103 loc) · 8.81 KB

Adjacencies.md

File metadata and controls

129 lines (103 loc) · 8.81 KB
Raw

Adjacencies

Contents:

Adjacencies overview

One of the key concepts in the RCK model is the notion of adjacency. Adjacency is a transition between segment's extremities. There are two kinds of adjacencies:

  • reference (present in the reference genome, or inherited by the derived cancer genome(s))
  • novel (present in derive genomes only).

Every adjacency {(q,x,+|-),(p,y,+|-)} describes a transition from (right|left) side of loci at coordinate x on chromosome q to (right|left) side of the loci at coordinate y on the chromosome p. Reference adjacencies naturally have a form of {(chr,x,+),(chr,x+1,-)} (i.e., on the same chromosome, neighbouring positions, and respective extremities via strands orientation).

We call two adjacencies reciprocal if some pair of their extremities resemble a reference adjacency. For example, adjacencies {(1,123450,+),(1,4567890,+)} and {(1,123451,-),(1,876534,+)} are reciprocal, because extremity (1,123450,+) form first adjacency one, and extremity (1,123451,-) from second adjacency resemble a reference adjacency {(1,123450,+),(1,123451,-)}

While sometimes novel adjacencies are classified as insertion, deletion, duplication, reversals (aka inversion), translocation, etc. This is usually done by looking at chromosomes, coordinates, and strands of involved extremities. For example insertion and deletion has the same signature (i.e., same chromosome, + strand on the leftmost extremity, and - on the rightmost extremity). Duplication has a signature of - strand followed by the - strand on the same chromosome. Reversal (event) usually involves two reciprocal novel adjacencies, with +,+ strand signature on one, and -, - signature on another adjacency.

While, indeed, aforementioned annotation correspond to cases, where respective rearrangement events would produce such novel adjacencies, it can also be the case that novel adjacencies that resemble signatures described above can be produced by more complex rearrangement events, such as chromoplexy and chromothripsis.

RCK adjacency format

RCK works with adjacencies in the following (T/C)SV (Tab/Comma Separated Values) text format:

aid	chr1	coord1	strand1	chr2	coord2	strand2	extra

where every entry thus describes an adjacency {(chr1, coord1, strand1), (chr2, coord2, strand2)} with an id of aid. The extra field is designed to store key=value pairs of additional information about adjacencies, with entries being separated by ;.

There are several special extra fields, that RCK relies upon, when working:

  • aid -- copy of the aid. When reading adjacencies, id for the adjacency will be based on the column, not the extra field.
  • cn -- copy number values (refer to the following subsection)
  • at -- adjacency type (either N for novel (default), or R for reference). By default all adjacencies are considered to be noevl, unless the adjacency id starts with the lower-case r.

inferred clone- and haplotype-specific adjacency copy numbers

The result sof the main RCK algorithm (via rck executable) contains the rck.acnt.tsv file, with entries following the RCK adjacencies format.

Both novel and reference adjacencies are output in the result, and depending on the --o-acnt-mix-novel-and-reference novel and reference adjacencies are either going to be mixed together, or separated with novel adjacencies followed by the reference ones. While the adjacencies themselves are self-explanatory, the main important peace of information about them is the cn field in the extra column, that encodes the clone- and haplotype-specific copy number values.

The cn value is a python/JSON dict with the following structure:

{
  'clone_id': {
    'AA': int,
    'AB': int,
    'BA': int,
    'BB': int
  },
  ...
}

where clone_id corresponds to the clone, for which haplotype-specific copy numbers are provided, and the AA, AB, BA, BB entries encode the copy number of the (haplotype) labeled versions of the adjacency (where the first position is labeled with the first haplotype letter, and the second position is labeled with the second haplotype letter).

In the following example:

aid	chr1	coord1	strand1	chr2	coord2	strand2	extra
id1	1	123450	+	1	123760	-	cn={'c1':{'AA': 1, 'AB': 0, 'BA': 0, 'BB':0}, 'c2': {'AA': 2, 'AB': 0, 'BA': 0, 'BB':0}}

where for the novel adjacency {(1,123450,+),(1,123760,-)} with id id1 the following labeled adjacency {(1,123450,+,A),(1,123760,-,A)} has a copy number 1 in clone c1 and copy number 2 in clone c2.

Converting to RCK format from SV detection tools

RCK installation adds rck-adj-x2rck adjacency-conversion executable tool to the PATH of your installation environment. With the help of rck-adj-x2rck one can convert (unlabeled) novel adjacency predictions from the following tools:

The input is converted to the RCK adjacency format (described above). rck-adj-x2rck tries to retain as much as possible of extra information from the SV detection tools input during the conversion, and such information is stored in the extra column.

For help message of rck-adj-x2rckrun:

rck-adj-x2rck --help

To get help in converting adjacency prediction from a specific tool x run:

rck-adj-x2rck x --help

The following optional command line arguments are shared for all of the input sources and can be beneficial to use:

  • -o | --output -- output file (default is stdout)
  • --id-suffix -- a suffix, that will be appended to every input adjacency id, when transforming to RCK format. This can be beneficial when working with several input sources of adjacencies, and one would want to differentiate based on the source

An example of converting adjacency prediction by Sniffles tool to the RCK suitable format:

rck-adj-x2rck sniffles SV_predictions.vcf --id-suffix sample-technology-sniffles -o sample-technology-sniffles.rck.adj.tsv

Which will convert SV prediction in SV_predictions.vcf produced by Sniffles in sample sample that was sequenced with technology technology into the RCK formatted adjacency calls in the file sample-technology-sniffles.rck.adj.tsv.

All converted adjacencies will have id_sample-technology-sniffles, where id is the VCF id provided by Sniffles.

Not that the --id-suffix value here is provided as an example and is not mandatory to (i) be prent at all (default value is empty string) (ii) be in the form of sample-technology-method, though we found it useful on several occasions. Depending on your needs a different suffix values may be more useful.

Processing RCK adjacencies

RCK installation adds rck-adj-process adjacency processing executable tool to PATH of your installation environment. For rck-adj-process the following commands are available:

  • cat -- combining adjacencies from 1+ inputs into a single one
  • reciprocal -- updating extremities of adjacencies in the input, so that pairs of extremities of distinct adjacencies that resemble reciprocality, but are tno exactly 1 bp apart, are brought together. This option ran by default in the main rck executable, unless explicitly suppressed.

Running rck-adj-process command --help provides one with the help on usage of each particular command.