2019-08-15 PFR-GOBii Collaboration; Polyploid support; Data storage solutions; intern.

2019-08-15 PFR-GOBii Collaboration; Polyploid support; Data storage solutions; intern.

Date

Aug 15, 2019

Participants

  • @Yaw Nti-Addae

  • @Elizabeth Jones

  • @Hymmi Kong (Deactivated)

  • @Chetan Baadkar (Deactivated)

  • @Tim Millar

  • @Susan Thomson

Goals

  •  

Discussion topics

Item

Notes

Item

Notes

Polyploid support

  • PFR still trying to figure out how to analyze data

  • PFR has no set format yet

  • Current have VCF files

  • Would like to store hyplotypes and polyploid genotypes

  • Will GOBii be the main repository for VCF data???

 

GOBii data storage solutions for polyploids/complex data types

  • Diploidize data

  • Add new HDF5 datatype for tetraploids etc – PFR - for SNPs only for now A/A/A/T and A|A|A|T optional vcf for ps (phase set) integer for group specific for a dataset (but would want to use consistently across all datasets) . Ps not urgent.

  • Padding HDF5 : AAAA and AA become AAAA and AA__

  • Look-up table : A / AAATA replaced ù / ĉ in HDF5, with translation back upon extract

  • Try PostgreSQL 11: can handle variable length alleles

  • Downstream analysis need read depth at diploid level, or dosage alleles

  • A/T:15,6 or AAAT or T:0.25 or 0/0/0/1; 0.25, 0.25, 0.25, 0.25

  • Filtered markers – remove from file – already removed from the vcf during cleaning phase

Student intern project: data management polyploids and indels

  • With simulated dataset 1250 samples, 0.5M markers, with mix of indels and ploidy

  • Padded to largest indel

  • Look up table for indels (and padding for ploidy)

  • PostgreSQL 11 with partitioning, parallelization and indexing

Questions

  • What is the genotyping data format of choice?

    • Diploidized data – don’t want

    • Nucleotides with dosage eg A/A/A/T - OK

    • Maximum likelihood dosage eg GT:AD:GP 0/0/0/1:13,6:0.033,0.684,0.233,0.045,0.005 – not sure

    • Haplotypes (1 row or multiple rows?) – multiple row easier ?

    • Raw vcf; standard or variable formats? May still be variable – would need postgres solution then

    • Geno call and vcf? Eg A/A/A/A; 0/0:3,0:3:88:0,9,108 – yes, Tim/Sue to send the best format

    • Are there any mixed ploidy files (variable length) - No

  • Indels? Max size? Proportion? –5% and low confidence, 2-10 bp multiallelic

  • What are the extract use-cases; cross ploidy, indels and SNPs, A/A/A/A and read depth format (Quality score too?)

Lee’s travel to Cornell

GOBii hackathon is Sep 23, 2019 to Sep 27, 2019 .

Action items

@Yaw Nti-Addae will email @Guy Davenport (Deactivated) and @Lee O'Grady with possible dates for travel to Cornell

Decisions