- Proposed a convex mixed integer quadratic program reformulation of the original non-convex problem by using integer programming techniques.
- Showed that the proposed reformulation can be efficiently solved with mixed integer programming solvers to identify strains from mixed diagnosis sample with up to three sub-types and 24 DNA measurement locations.
- Showed that the reformulation can also be used to study the ambiguity of the inverse problem.
More details can be found in the paper:
Lauri Mustonen, Xiangxi Gao, Asteroide Santana, Rebecca Mitchell, Lars Ruthotto, Ymir Vigfusson. A Bayesian framework for strain identification from mixed diagnostic sample, Inverse Problems, 2017.