CrossStitch: Hybrid Phasing and Personal Genome Construction

CrossStitch creates personalized reference-quality diploid genomes without de novo assembly. The basic idea is rather than trying to assemble a genome from scratch, it will leverage a reference genome as a baseline, and then update it with any SNPs, indels, or structural variations present in your sample. For the best results, the data requirements are similar to a de novo assembly: Illumina-based data for SNPs and Indels, Long Read data for structural variants, and Phasing data such as 10X Linked Reads and/or HiC data. However the CrossStitch process is much less demanding, produces more accurate results, and the process is much more predictable. The output will be a phased VCF file with all variants (SNPs, Indels, and SVs) as well as a phased personalized diploid genome including 2 copies of each chromosome with the variants inserted at the correct locations.

Installation

## CrossStitch requires extractHairs from HapCut2
$ git clone https://github.com/vibansal/HapCUT2
$ cd HapCUT2
$ make

$ git clone --recursive https://github.com/schatzlab/crossstitch.git
$ cd vcf2diploid
$ make
$ cd ../Iris
$ build.sh

## After this you will need to add HapCUT2/build/extractHAIRS to your path or edit src/crossstitch.sh with the correct path

Running CrossStitch

Currently only human genomes are supported for diploid genome construction.

$ crossstitch.sh phased_snps.vcf unphased_structural_variants.vcf long_reads.bam genome.fa outputprefix gender refine
 
Details:
  phased_snps.vcf:                   VCF file of phased SNP and indel variants. Recommend LongRanger (10X only) or HapCUT2 (HiC and/or 10X)
  unphased_structural_variants.vcf:  VCF file of structural variants identified using Sniffles
  long_reads.bam:                    BAM file of long reads aligned with NGMLR
  genome.fa:                         Reference genome used
  outputprefix:                      Prefix for output files
  karyotype:                         "xy" or "xx", used to ensure sex chromosomes are correctly used
  refine:                            optionally refine structural variant calls with local assembly (1=refine, 0=skip)

Running Insertion Refinement

The unphased structural variant must include the names of supporting reads for insertions (sniffles -n [x], where x > 0).

To run the insertion refinement on its own, use the Iris submodule. It has dependencies on minimap2, racon, and samtools.

Simulations for testing

Simple example: 5 indels, zero other SVs

$ cd crossstitch/simul/simple
$ ../sim_example.sh 1

Complex example: 20 indels, 10 inversions

$ cd crossstitch/simul/fiftyinv
$ ../sim_example.sh 1

Creating you own simulations

Create/copy the simul.param file from the simple example with more/less SVs of different types. Then run ../sim_example.sh

Output files

Main Files:

*.hap1.fa.gz: Haplotype 1 chromosome fasta sequences 
*.hap2.fa.gz: Haplotype 2 chromosome fasts sequences 
*.spliced.scrubbed.vcf.gz: Finalized set of phased small and structural variants 

Annotation files:

*.map:   liftover file to relate coordinates on the personalized assembly to the reference (such as GRCh38) 
*.chain: liftover file to relate coordinates on the personalized assembly to the reference (such as GRCh38)