number of trans-qtl tests

[liaguiro]

Hello,

I calculated about 24.8 M possible trans qtl tests but got 18.8 M trans-qtl test results when I ran the trans-QTL mapping with --pval_threshold 1. I'm wondering why tensor-qtl ran fewer trans-qtl tests. Does tensorqtl do additional filtering? or did I estimate the number of possible trans-qtl tests wrong?

same.chr.cis.gene.snp <- unnest(gene.TSS, col = snp.table) %>% mutate(same.chr = gene.Chr == snp.CHROM, dist.gene.snp.window = snp.POS >= gene.TSS.start - 500000 & snp.POS <= gene.TSS.end + 500000) %>% filter(same.chr) %>% filter(dist.gene.snp.window)

same.chr.trans.gene.snp <- unnest(gene.TSS, col = snp.table) %>% mutate(same.chr = gene.Chr == snp.CHROM, dist.gene.snp.window = snp.POS >= gene.TSS.start - 500000 & snp.POS <= gene.TSS.end + 500000) %>% filter(same.chr) %>% filter(!dist.gene.snp.window)

dif.chr.trans.gene.snp <- unnest(gene.TSS, col = snp.table) %>% mutate(same.chr = gene.Chr == snp.CHROM, dist.gene.snp.window = snp.POS >= gene.TSS.start - 500000 & snp.POS <= gene.TSS.end + 500000) %>% filter(!same.chr)

all.trans.gene.snp <- bind_rows(same.chr.trans.gene.snp, dif.chr.trans.gene.snp)

[francois-a]

The default maf_threshold for trans-QTL mapping is 0.05, which likely explains the missing pairs. If you want to output all pairs, you may want to use --return_dense instead of setting pval_threshold to 1.

[liaguiro]

I used --return_dense instead of --pval_threshold to 1 and got the following error. When I ran trans-qtl with pval_treshold, it worked. I'm not sure how to fix this error

/tensorqtl/trans.py:188: FutureWarning: Series.getitem treating keys as positions is deprecated. In a future version, integer keys will always be treated as labels (consistent with DataFrame behavior). To access a value by position, use ser.iloc[pos]
interaction_mask_t = torch.BoolTensor(mask_s).to(device)
trans-QTL mapping

• 287 samples
• 22447 phenotypes
• 1 covariates
• 1111 variants
• including interaction term
  ^M processing batch 1/1
  Traceback (most recent call last):
  File "", line 198, in _run_module_as_main
  File "", line 88, in _run_code
  File "/tensorqtl/main.py", line 2, in
  tensorqtl.main()
  File "/tensorqtl/tensorqtl.py", line 328, in main
  pairs_df = trans.map_trans(genotype_df, phenotype_df, covariates_df=covariates_df, interaction_s=interaction_df,
  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/tensorqtl/trans.py", line 276, in map_trans
  res = calculate_interaction_nominal(genotypes_t, phenotypes_t, interaction_t, residualizer,
  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/tensorqtl/core.py", line 189, in calculate_interaction_nominal
  i0_t = residualizer.transform(i0_t.t(), center=False).t()
  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/tensorqtl/core.py", line 62, in transform
  M0_t = M_t - torch.mm(torch.mm(M0_t, self.Q_t), self.Q_t.t())
  ^^^^^^^^^^^^^^^^^^^^^^^^
  RuntimeError: mat1 and mat2 shapes cannot be multiplied (287x1 and 287x1)