abayega / microbe

A deep learning model to remove batch effects for gut microbiome data obtained from IBD patients.

microBE

A deep learning model to classify between microbiota of healthy and IBD patients.

Preprocessing the data

The scripts used to preprocess data with pointwise mutual information and centered log-ratio (CLR) normalization can be found in /preprocessing/Data_preprocessing_functions.py. Three scaling methods were explored: standardization (using mean and standard deviation), min-max, and median absolute deviation (MAD), which can be found in /preprocessing/normalization.py (converted from .ipynb to .py).

Classifers

Several traditional classifiers as well as VAEs were implemented to handle the healthy/diseased classification as well as batch effects. They can be found under /classifiers.

• Logistic Regression
• SVM (Linear and RBF)
• RF
• MLP
• DANN (Domain Adaptation Neural Net)
• DeepMicro (modified DM.py file for predefined training and test sets, for more information on how to implement: https://github.com/minoh0201/DeepMicro)

System Diagram

System Diagram - DANN