drastic

Deep Rapid Annotation using Submodels Training in Cells

Table of contents

• What is drastic?
• Getting started
• Features
  • Further in the horizon
• Command Line Interface
• Schema of the model
• What We Learned
• Refereces

What is drastic?

Drastic provides a Deep Learning model to perform annotation of a prokaryotic genome. This project is highly inspired by DeepAnnotator [1] (repository here).

Getting started

How to clone, add bin to the PATH and use the model to train a model here.

Features

This project aims to provide a deep neural network capable of annotating a prokaryotic genome.

• Trained LSTM neural network model to find genes in user provided sequences.
• Embeddings for the NLP problem.
• Score performance based on [1].
• Trained LSTM neural network model to define start and end of genes in user provided genome.
• Trained LSTM neural network to define start and end of genes and also protein coding sequence.
• Treating of edge-cases (rRNA, tRNA, CRIPRs...).

Further in the horizon

Alternatively, the user could provide its own data to train the model.

• Profiles of sequences.
• CLI interface.
• Pre-process data supplied by the user.
• Train the model with this data.
• Use the model to predict.
• Streamlit interface.

Command Line Interface

Provided as help, in this section, the usage of the model should be summarized.

Schema of the model

An explanation (better with a graph) should be placed here.

What We Learned

This project was developed for the MSc course "Deep Learning" of the Technical University of Denmark.

References

[1] Amin, M. R., Yurovsky, A., Tian, Y., & Skiena, S. (2018). DeepAnnotator. The 2018 ACM International Conference