Output schemas¤
Mapping outputs are compressed Parquet files. Columns specific to the Negative Binomial model or permutation calibration are omitted when they do not apply.
Cis output¤
jaxqtl cis writes ${out}.cis.${test}.${perm|acat}.parquet.gz.
| Group | Columns |
|---|---|
| Phenotype | phenotype_id, chrom, num_var |
| Lead variant | snp, a1, a0, pos, tss_distance, af, ma_count |
| Association | beta, se, pvalue, pvalue_adj, adj_method |
| Model | nb_alpha, negloglikelihood, model_converged |
| Validity | result_valid, failure_reason |
| Beta calibration | shape1, shape2, nc_estimate, perm_converged |
The Beta-calibration fields are not emitted for ACAT. nb_alpha is not emitted for Gaussian or Poisson models.
negloglikelihood is the fitted negative log-likelihood objective (lower is better). Score and SPA tests report the
shared null-model objective; Wald tests report the selected variant's full-model objective.
Invalid rows preserve the tested phenotype
A gene with no finite SNP-level p-values remains in the cis output with result_valid = false. Its lead variant,
association statistics, and convergence values are null. Use failure_reason to distinguish this state from a
valid association with a large p-value.
Nominal output¤
jaxqtl nominal writes ${out}.nominal.${test}.parquet.gz with one row per phenotype–variant pair:
phenotype_id,chrom,snp,pos,a1,a0,tss_distance,af,ma_count.beta,se,pvalue,negloglikelihood,model_converged.nb_alphafor Negative Binomial models.
For score and SPA tests, negloglikelihood is shared across all variants for a phenotype because they use one fitted
null model. For Wald tests, it is the fitted full-model objective for that phenotype–variant pair.
Trans output¤
jaxqtl trans writes two files:
${out}.trans.${test}.variant.info.parquet.gzcontainschrom,snp,pos,a1,a0,af, andma_count.${out}.trans.${test}.sumstats.parquet.gzcontainsphenotype,snp,beta,se,pvalue,model_converged, andnb_alphafor Negative Binomial models.
Within each phenotype block, summary-statistics rows follow the variant order in the metadata file.