Run eQTL analysis with TensorQTL

[lahuuki]

TensorQTL: https://github.com/broadinstitute/tensorqtl (Check documentation)

Code from MDDseq: https://github.com/LieberInstitute/goesHyde_mdd_rnaseq/tree/master/eqtl/code

Data conversion

Covariates

• covariate data: model + gene PCs
• feature PCs: TODO check if already done in Hb
• what model to use? (simple for MDDseq: mod)
  • formatting covariate tables: covar_format()

Expression

• run all four features ? Or just Gene?
• rse_to_bed : note start/end handling

Genotype

• /dcs04/lieber/lcolladotor/pilotHb_LIBD001/Roche_Habenula/processed-data/08_bulk_snpPC/08_bulk_snpPC/habenula_genotypes.* (bed, bim, fam, bam files are on JHPCE)

Run tensorQTL

• Do we want to run just Nominal? or cis/independent too?
• nominal run
• my_tensorqtl_run.py : helper functions to read in data to tensorQTL, check genotype names, add 'chr' to genotype data, rm missing chrs from expression data

p-val correction -04_filter_genomewide_nominal.R : Read parquests back in to R, combine all chrs, calc FDR, filter and write csv with FDR> 0.05

[lahuuki]

Adding a cell type interaction to nominal run: https://github.com/LieberInstitute/goesHyde_mdd_rnaseq/blob/d8a3876e941f0b38791b5dfc1138f536d0e86e0a/eqtl/code/09_tensorqtl_genomewide_interaction.py#L19C8-L29C53