Kimmo Palin

Second this and would implement is as bed file for exclusion regions. Also reporting a documentation bug: --problematic parameter takes a bed file, not "pipe-delimited list of regions"

Haven't tried. That should work but would smoothen the gc percentages fairly drastically. My current plan for workaround is to work with goleft indexcov https://github.com/brentp/goleft Valeriu Ohan kirjoitti ti 19....

It is a bit of a pain to adjust the few troublesome SNPs (12 at this case) but here it is [sites.chm13v2.rna.vcf.gz](https://github.com/brentp/somalier/files/9980280/sites.chm13v2.rna.vcf.gz) The troublesome sites (deleted in T2T assembly) are...

Yes I'd like to have a predictor, e.g. ``` with pm.Model(): w_prior = pm.Dirichlet("w", a=[1, 1]) beta = pm.NormalMixture("beta", w=w_prior, mu=[0, 0], sigma=[0.1, 4.5], shape=D) sigma = pm.HalfStudentT("sigma", nu=4, sigma=2.0322)...

That seems to work, also with formula like `response ~ 0+d0+d1+d2+d3+d4+d5+d6+d7+d8+d9` (I've never before seen the `c(d0,d1...)` syntax.) I don't have proper references for this apart from the blog post...