INSTRUCTION MANUAL

1. Create the new reference (ref+), which includes the structural variations of interest. 1.1 reference_generation/run.sh executes the program for reference generation. 1.2 The program creates a new reference from an existing one by incorporating structural variants from a file provided by the user. The variation types are INSERTION, DELETION, INVERSION, TANDEM, TRANSLOCATION,
   and SNP. The TRANSLOCATION and DELETION types actually remove sequence content from the original refernce. This is usually not an intended effect, and better results are probably achieved by replacing TRANSLOCATIONs with sequence duplications (INSERTION) and ommitting DELETIONs altogether. These types are included for completeness only. 1.3 The variation type should have the following format:
   TYPE LOCUS [SEQUENCE/LOCUS2] The sequence is optional for those the types requiring additional information of the inserted sequence. 1.4 Examples: INSERTION chr1:1024878-1024878 gaggttaaaatctcc DELETION chr2:2017674-2018635 SNP chr3:1288100-1288100 T TRANSLOCATION chr4:2388190-2388190 chr5:4377179-4377817

2. Index the new reference, and align reads to it with a mapper of choice (bwa, bowtie2, ...) -- the mapper is best chosen depending on the SV detection tools used in the next step.

3. Run an SV detection tool on the aligned read set (Delly, BreakDancer, Socrates). The choice here is dependent on the properties of the sequencing library: soft-clip based tools such as Socrates require long reads and deep coverage; pure paired-end methods such as BreakDancer rely on high average insert size and a tight distribution for maximum effct. Multiple tools can be run at this step if desired.

4. Evaluate the results. To assess the presence or absence of the variants added to the novel reference, the output from the SV detection has to be compared to the list of variants. A simple overlapping tool in results_evaluation takes two bed files and establishes the recall, precision, and f-score of the best overlap of intervals in the first bed file and any interval in the second bed file. 4.1 The interval_overlap_deletions.py script takes two files as an input. In the context of the ref+ pipeline, these files should be the list of variants and the (parsed) output of the SV caller. Both files are expected to be in standard tab- or whitespace-delimited format with each row containing the reference chromosome, the starting position, and the end position. Neither of the files have to be sorted, but can be arranged in any order. This makes the script slow and inefficient, but relieves some of the strains of pre-processing. The output of the script contains the variation coordinates, followed by three evaluation metrics, and the line of output from the SV caller. The metrics are the recall, the precision, and the f-score for the variant from file1 and the best hit from file2. If there are no overlapping intervals for a variant in file1, the metrics are set to 0. Otherwise, recall describes the proportion of the variant being overlapped by a call in file2, precision the proportion of the overlapping SV call, and F-score the combination of the two (2recallprecision)/(recall+precision).