Genomics Workflows🔗

IDAT to raw PLINK files🔗

• Raw IDAT generated by vendor and placed within HST secure computing environment
• Workflow run which takes raw IDAT files:
  • Normalization
  • Genotyping
  • Final report
  • PLINK conversion

PLINK Files🔗

PLINK files: .bed, .bim, .fam - Space delimited with six columns:🔗

1. Family ID (FID)
2. Within-family ID (IID) - i.e. FID plus relationship in family C = child, M = mother
3. Within-family ID of father
4. Within-family ID of mother
5. Sex
6. Phenotype

batch.info - Tab delimited with 3 columns:🔗

1. IID (character, 10 digits + 1 letter either M or C)
2. plate_number (float)
3. visit (character)

.fam and batch.info have the same ordering of IID column🔗

Genomic Derivatives🔗

• Principal components - .eigenvec, .eigenval
• Inferred Sex
• Inferred Ancestry

Prerelease staging file structure🔗

• Example directory structure for internal data storage
  • This data is not ready for public as it is not deidentified!!!
  • Only files in the “data” subdirectory are to be shared other files are supporting files in generating the release data
• For each release files will be placed here before running deidentification
  • /projects/standard/basu_hbcd/shared/HBCD_genomics_release_<release details>
  • Newly associated files that are not included release (such as par_visit_data.tsv) will be in the firs level of this directory
  • Data to be released will be in the “data” directory
  • QC’d genomic data will be stored as Plink1 binary (.bed/.bim/.fam) format in the first level of the “data” directory
  • Derivatives associated with genomics release will be stored in “data/derivatives”

Generation of release ready PLINK files🔗

• Sample-Level QC (match with .fam file generated by Sampled) - call rate, sex check, family relatedness, PCA
• Sample-Level QC (match with the LORIS demographic file) - sex check, relatedness, ancestry
• Variant level QC - Batch effects
• To ensure that PLINKs are ready for a given release they must be filtered with the par_visit_data.tsv used for the release

Deidentification, folder structure, and push to release bucket🔗

• Use a recent mapping file present within s3://midb-hbcd-main-pr-deidentification-list - maps pscid to release candid
• Script within https://github.umn.edu/MIDB-IG/data-release/tree/genomics-de-id named deidentify_genomics.py
  • Will de-identify .fam, and batch info file based on provided mapping file
  • Copy de-identified .fam and batch info file along with .bed and .bim files into a separate folder
  • Folder designed to emulate ABCD directory structure (i.e. <output folder>/concat/genetics/genotype_microarray/GDA)
• Once produced a validation script will be run to ensure that appropriate participants are present and that all are de-id’ed
  • Confirming that FID columns are completely numeric
  • Confirming that IID columns are numeric plus a single character (1 of either “M” or “C”)
  • Confirming that the FID and IID columns are of fixed width (9 and 10 characters respectively
• Once validated, de-id’ed release data will be pushed to pre-release bucket - i.e.BR: s3://midb-hbcd-lasso-hdcc-qc-br/br<number>