About MegaD

MegaD is a new iteration of previous project MegaR which aims to utilize deep neural networks for disease status classification based on taxonomic profiles derived from metagenomic sequence data. MegaD differs from MegaR in a few key ways. Most notably, a change from R to Python, integration of neural networks, and providing the user with a set of pre-trained models for quick classification of unknown samples.

The workflow of MegaD from sequence to prediction is as follows:

1. Raw sequences are derived from 16S sequencing or Whole Genome sequencing
2. Taxonomic profiles are created using either the provided script or manually through the use of software such as Kraken2 or MetAPhlAn3.
3. a. A model is trained using a set of training data for the specific disease being tested for. This model can be saved for future use.
4. b. Alternatively, one may choose to use of the provided pre-trained models
5. The trained model is used to predict the disease status of an unknown sample.