Advanced usage

Running sub-models

By default, the run_model function uses the full HAMdetector model that includes the horseshoe prior, phylogeny information and epitope prediction. Reduced versions of the full model can be run by supplying an additional positional argument:

using Escape

data = HLAData(
    alignment_file = "/home/user/Desktop/alignment.fasta",
    tree_file = "/home/user/Desktop/phylogeny.tree"
)

result = Escape.run_model(HLAModel{4}(), data)

The different versions are:

• 1: A simple logistic regression model
• 2: Like model 1, but additionally with horseshoe prior
• 3: Like model 2, but additionally with phylogeny information included
• 4: Full HAMdetector model, like model 3 but with epitope prediction included.

Note that the full HAMdetector model includes all the previous models as a special case.

Modifying sampler parameters

By default, 4 chains with 1000 iterations each and a warmup of 300 samples are drawn. This should be plenty for all use-cases, but can be changed as shown below:

using Escape

data = HLAData(
    alignment_file = "/home/user/Desktop/alignment.fasta",
    tree_file = "/home/user/Desktop/phylogeny.tree"
)

result = Escape.run_model(
    data, chains = 4, iter = 1000, warmup = 300, stan_args = "adapt delta=0.95"
)

For a full list of the CmdStan options, please refer to the CmdStan sampling parameters. It should not be necessary to modify the stan_args argument, it is just listed here for completeness.

Using 4-digit HLA alleles

By default, HAMdetector uses HLA alleles with 2-digit accuracy. To use HLA alleles with 4-digit accuracy (if supplied), set the allele_depth keyword argument to 2 (default is 1).

using Escape

data = HLAData(
    alignment_file = "/home/user/Desktop/alignment_with_4_digit_alleles.fasta",
    tree_file = "/home/user/Desktop/phylogeny.tree",
    allele_depth = 2
)