A class of Directional Guassian Mixture Models (GMMs) for recovering microbial spatial structre from MaPS-seq data. Data is from Sheth et al.
environment.yml contains the virtual environment that was used to generate our results.
Data from the sm1 experiment in Sheth et al is filtered as described in the manuscript. sml_data.p contains the filtered data.
basic_GMM_EM.py contains an implemention of a standard Expectation Maximization (EM) algorithm for parameter inference in a naive GMM. EM_intercept_inference.py and EM_twodirection_inference.py contains the EM algorithms for the single- and two-directional GMMs respectively. Mathematical proofs of the accuracy of our algorithms can be provided upon request.
model_selection_sm1.py provides an implemention for loading the filtered sm1 data and finding the best fitted model. We have provided the model selection results for the Cecum dataset.