Question about sequence order using a software tool

Question

[mangfu100]

Hi

I have a question when using software tools SVDetect.

SVDetect finds structural rearrangments by classing subtypes such as translocation, inversion duplication and etc.

Anyway, I don't understand the meaning of order filtering. I guess that it might be a kind of filtering process which is using order of reads. And order of reads is specified in SAM format. so I look SAM files and got the information of reads but can't figure out the real meaning in SVDetect order filtering.

Simple explanation is specified in homepage and below is that .

-- order_filtering(o) flag to run the order filtering
-boolean, value= 1 (run), 0 (skip) -- nb_pairs_order_threshold(o) minimal number of pairs in a subgroup of paired-end reads for balanced events (order_filtering only). Also used to detect balanced rearrangements -integer, ex: 2 (default=1)

Also , I open the perl files and look the codes. But it was really hard working.. and also I don't understand meaning of order filtering.

I am looking forward to your reply.

Thank you.

Answer 1

[cclementddel]

Did you find something, almost one year ago ? :/

Here is explanations you can find in supplementary data :

Order filtering algorithm. When the order filtering algorithm is called, the information of the cluster order, direct or inverted, is known. For each pair P, we can find (i) pairs that are closely mapped to P, (ii) pairs which are located at a larger distance to P and whose order is in agreement with P, and (iii) pairs located at a larger distance to P and whose order disagrees with P. If there are no pairs in the cluster whose orders disagree with each other, the whole read cluster is conserved, and the order filtering is stopped. Otherwise, we continue the algorithm by removing the read pair showing the highest number of cases (iii). Then, we check again if all pair orders are in agreement and if it is still not the case, the next pair is removed following the same decision parameters. We iterate until the order of all pairs is correct.

In the output file of SVDetect, we indicate how many read pairs have been removed after running this procedure. The user can then define a cutoff based on the percentage of the pairs that have been filtered out. We suggest keeping clusters which had, at most, 1/3 of read pairs discarded.