This repo contains a partial Python implementation of the STRUCTURE algorithm, completed as a required final project for CS4775, taken at Cornell University in Fall 2016. The primary motivation for this project was to implement a non-trivial approach to a problem in the field of computational biology. The objectives were to implement an approach discussed in the literature and apply it to a real dataset.
The implementation can be found in structureJonathan.py. The repository also contains the source code for the C implementation of STRUCTURE, which can be downloaded here. A PDF writeup of my results can be found in structureWriteup.pdf.
In the field of population genetics, classification of individuals into populations is a common method used to study various problems. STRUCTURE takes a model-based clustering method for using multilocus genotype data to infer structure and assign individuals to populations. More information can be found in the writeup or on the official STRUCTURE website here.
The algorithms are considered on a portion of the Hapmap3 dataset, as acquired by Admixture.
At the time of completion, the model without admixture was fully implemented, showing good results with both STRUCTURE and Admixture. The model with admixture is mostly implemented, but does not show results that agree with STRUCTURE or Admixture. I estimate that the implementation is 90% complete.
I, Jonathan, am the sole contributor to both the implementation and data analysis. Guidance was generously given by Professor Williams and Melissa Hubisz.