inheritance models for mendelian diseases

This module is a general-purpose framework for evaluating if a family exihibits, for example, and autosomal dominant pattern. The logic for this was tuned in gemini but we make it available here as a more general purpose library to encourage:

1. community driven improvements
2. use outside of gemini
3. more comprehensive testing

Finding variants that match autosomal dominance in a trio, for example is very simple to find, however, after considering multiple generations, arbitrary family sizes, depth cutoffs, and unknown genotypes and phenotypes to support to real-world datasets it becomes tedious and error-prone.

Supported inheritance tests

• autosomal dominant
• autosomal recessive
• de novo
• X-linked dominant, recessive, and de novo
• compound heterozygote
• mendelian violation

Usage

For now, the use is via api only. Users can look at the tests to see how to use. Most functions have a signature like:

 auto_dom(self, min_depth=0, gt_ll=False, strict=True, only_affected=True)

where the arguments enforce a minimum depth, a maximum genotype likelihood, strictness (mostly related to parent-offspring requirements) and wether to allow unaffecteds to have the variant (or be homozygous alt).

ToDo

1. add support for X-linked soon.
2. add a simple example of running on a VCF+PED
3. code coverage
4. code documentation

Testing

Tests can be run as:

nosetests --with-coverage -x --with-doctest --cover-package inheritance

Overview

the generic code is in inheritance/inheritance.py and a specific implementation that we use in gemini is in inheritance/pyeval.py. To make this available for a new resource, for example to bcftools we would look at the implementation of Filter in inheritance/pyval.py and make the generated strings match those expected by bcftools.