Prohlatype: A Probabilistic Framework for HLA Typing

HLA typing from sequencing data is considered as a classical probabilistic inference problem and Profile Hidden Markov Models (PHMM) are motivated for the likelihood calculation. Their generative property makes them a natural and highly discernible method; at the cost of considerable computation. We discuss ways to ameliorate this burden, and present an implementation https://github.com/hammerlab/prohlatype.