SVsig - Recurrent structural variations detection in cancer whole genomes

SVsig is a method developed to classify rearrangements as passenger or driver in cancer patient cohort of whole genome sequences. The distribution of rearrangements in the cancer genome is shaped by both the mechanisms of their formation and the fitness advantages they confer on the cell. This analysis revealed significant predictors of the distribution of rearrangement across the genome and identified known and novel rearrangements that recurred more often than expected given these predictions (for more detailed description: https://doi.org/10.1101/2023.10.13.561748)

MATLAB toolboxes needed:

• Statistics and Machine Learning Toolbox
• Optimization Toolbox

This version has primarily been tested using MATLAB_R2020a.

How to run SVsig

Preparing files

SVSig takes in an input .csv file with 15 columns. An example is in this repo under data/merged_1.6.1.csv. Your file must match the column names exactly.

• seqnames, start, strand, altchr, altpos, altstrand: genomic coordinates of both rearrangement breakpoints.
  • Note that chromosome coordinates are integers only. chrX and chrY are changed to 23 and 24, respectively.
• dcc_project_code: histology or tissue type information.
• sv_id: ID for individual rearrangement
• sid: Patient ID for the rearrangement
• donor_unique_id: Patient ID for the rearrangement
• weights: Ranges from 0-1, representing the weight an individual connection is given to an entire rearragnement event.

Optional columns: If this information is not available, set column values to arbitrary value. Will not affect ability to run the model.

• topo: rearrangement topology information
• topo_n: number of rearrangements involved in topology.
• mech: DNA damage repair mechanism predicted to generate the rearrangement.
• homseq: number of base pairs of microhomology at the breakpoint junctions.

Simple Rearrangements Model (SVsig-2D)

SVsig-2D considers each rearrangement to occur independently of each other.

• Open runSVsig.m
  • Change path to sample rearrangements file within lines 22-43
• Open Run2DModel.m
  • Make sure: complex, weights, and simulation parameters are all false
  • Set working directory in line 48
  • Change path to write hits table in line 83
  • Run Run2DModel.m

Complex Rearrangements Model (SVsig-2Dc)

• Open runSVsig.m
  • Change path to sample rearrangements file within lines 22-43
• Open Run2DModel.m
  • Set weights and complex parameters to true.
  • Set working directory in line 48
  • Change path to write hits table in line 83
  • Run Run2DModel.m

There are additional parameters that can be set in Run2DModel.m

• model_exist: Boolean to skip model training and use a pre-determined background model. If True, add path to background model in line XX of runSVSig.m
• len_filter: Only considers rearrangements above this length for calculating significance.
• bks_cluster: Set to 1.
• FDR_THRESHOLD: FDR threshold for determining significance.
• output_file: path to output file
• complex: Boolean to run SVSig-2Dc (complex model).
• num_breakpoints_per_bin: Average number of breakpoints within a bin. To determine bin boundaries. Currently not used. Note to self -- fix this later.
• bin_length: Length of bin to divide genome. Default: 500kb
• weights: Weight given to each individual connection, ranges from 0-1. Weight=1 for the simple model. For the complex model, weights are obtained from the juxtapositions file after running JaBbA
• simulations: Boolean to test simulated data.
• genome_build: 'hg19' or 'hg_38'.

Note to self: remove local and simulations parameter, add output_file parameter ALSO move parameters from break_invasion_model to Run2DModel

Outputs

SVsig-2D and SVsig-2Dc output a file containing significantly recurrently events. Each unique event is denoted with by a cluster number. The genomic coordinates, subtype, and ID information for each rearrangement in a cluster are displayed. In addition, the following columns are present:

• cluster_num: Cluster number each connection belongs to.
• pval: Significance for the rearrangement event.
• prob:
• num_hits: Number of unique samples containing the rearrangement.

Tutorial

License

Author: Shu Zhang, [email protected], [email protected]

Contact: Rameen Beroukhim, [email protected]

License: GNU AGPL3, Copyright (C) 2023 Dana-Farber Cancer Institute

Please cite: Zhang S, Kumar KH, Shapira O, et al. Detecting significantly recurrent genomic connections from simple and complex rearrangements in the cancer genome. bioRxiv (2023). https://doi.org/10.1101/2023.10.13.561748