VERSE: a novel approach to detect virus integration in host genomes through reference genome customization

VERSE: a novel approach to detect virus integration in host genomes through reference genome customization

Wang et al, Genome Medicine (2015)

Ino de Bruijn / @inodb

VirusFinder 1.0 (2013)

• detect viruses of any type
• integrated/unintegrated
• detect virus integration sites
• WGS, RNA-Seq, TS (paired end reads)

VirusFinder 1.0 (2013)

VERSE aka VirusFinder 2.0 (2015)

• Tries to fix
  • host genomic instability
  • high virus mutation rates
• Host and virus genome customization with iCORN
  • (Otto et al, 2010)
  • Iterative Correction of Reference Nucleotides
• iCORN detects SNPs and indels
  • adjust references iteratively
  • increase coverage over each base

VERSE (2015)

VERSE (2015)

• Two classes of predicted integration sites
  • high confidence: sufficient soft-clipped reads for CREST
  • low confidence: locus within region predicted by SVDetect that has most soft-clipped reads
    • distance between regions at least 10 bp

Benchmark

• WGS of 13 hepatocellular carcinomas (HCC)
  • Illumina 2x 90bp
  • mean cov 31.7x to 121.2x
  • 20 HBV validated integration sites (PCR, Sanger)
    • disregard 2 closer than 10 bp
• RNAseq of 4 HCC cell lines
  • Illumina 2x 101 bp
  • mean 127M reads per sample
  • 11 HBV validated chimeric transcripts (Sanger)

Benchmark

• TS of 2 Merkel cell carcinomas
  • 2x 100 bp
  • 3.9M and 5M reads
• Simulated data
  • Illumina 2x 75 bp
  • mean cov 30x
  • mutated HPV-16 + hg19
    • SNPs, indels, SVs

Results

• Virus genome customization
  • Simulated data
  • 91% SNPs, 67% indel 1 iteration
  • 99.9% identity with consensus HPV-16
    • compared to 99.1% reference

Results

Host genome customization

Sample 26T from WGS HCC

Virus integration site detection

State of the software

• Difficult to install
  • many dependencies are old versions
• The code is not in an online code repository
  • Difficult to propose changes
• After getting help, still a lot of warnings when running the software on a test set
• Excellent support right here at MSKCC

Comments

• Host/viral genome customization potential problems not discussed
  • What is theoretical best iCORN can do?
• Viral integration sites
  • Are found sites 1 bp resolution?
  • Only sensitivity?
  • What proportion False Positive and False Negative
  • Comparison low confidence high confidence cases
• Only paired end reads, long single would be useful
• Multiple virus detection is not automatic

Questions?