Returning beta estimates for multiple covariates


I am running a simple burden test using hl.logistic_regression_rows() on a matrix table mt_burden containing my genomic and covariate data. mt_burden contains the following:

row fields:
gene_symbol: str

column fields:
pheno.is_case: int32 (coded 0/1),
pheno.age_at_enrollment: int32,
pheno.pca_features: array (list of 10 PCs)

ind_het: bool (carrier status)

The following call runs as expected for a simple burden test without an interaction term, returning the effect size estimates for the ind_het covariate for each gene of interest in gene_symbol:

    covariates = [1.0, mt_burden.pheno.age_at_enrollment, 
        mt_burden.pca_features[1], ..., mt_burden.pca_features[10]]

    log_reg = hl.logistic_regression_rows(
        test = 'wald',
        y = mt_burden.pheno.is_case,
        x = mt_burden.ind_het,
        covariates = covariates,
        max_iterations = 50

I would like to run a similar analysis, but including a age\_at\_enrollment * ind\_het interaction term as a covariate. I can form this with the following…

mt_burden = mt_burden.annotate_entries(het_times_age = mt_burden.ind_het * mt_burden.pheno.age_at_enrollment)

…but then am running into a few problems:

  1. Since het_times_age is an entry field, hl.logistic_regression_rows is throwing an error when this is included in the list of covariates. I see this thread, but was wondering if there was any update on this functionality!
  2. More generally, I would sometimes like the logistic regression to return beta estimates for more than one covariate. I see that hl.logistic_regression_rows only accepts a single <float64> value for x, and it throws an error when I try and pass a list of covariates. Is there a way of accomplishing this?

Sorry for the trouble and thanks so much for your help!