ultra-chrome

Repository for the Human Data Analytics Project 2020/2021.

Project structure

/dataset (to download)
    /data
    - data.tar
    - data.tar.gz
/demo
/models
    /Base_NN
    /ChromeR
    /Deep_Chrome
    /separated (contains all the models trained individually)
    /Ultra_Chrome
/notebooks
    - attentive-chrome.ipynb
    - converter.ipynb
    - models.ipynb
    - parser.ipynb
    - visualize.ipynb
/papers
/report
    /assets
    - latex-project.zip
    - Report.pdf
/results

Dataset

The full dataset used in this project can be found here.

Demo

The demo was develop as an Angular webapp and is hosted by Firebase Hosting.

Live demo here.

Report

In the report folder you can find the final Report.pdf along with the latex project I developed on overleaf, an online latex environment.

Notebooks

All the code written can be found in the notebooks folder in jupyter notebook format.

In order to reproduce the results reported in the Report:

1. Download the dataset into the dataset folder

2. Run parser notebook: loads raw data and organizes into numpy array for future processing

3. Run models notebook: loads the parsed dataset, trains and saves the models, computes predictions and saves results

4. Run attentive-chrome notebook: download the pre-trained models of AttentiveChrome and computes the predictions for every cell type of the dataset

5. Run converter notebook: converts the tensorflow saved models into javascript compatible files ready to be used inside the demo webapp

6. Run visualize notebook: loads results and compares them with different plots

Useful links

• DeepChrome

• DeepChrome- deep-learning for predicting gene expression from histone modifications

• AttentiveChrome-Deep Attention Model to Understand Gene Regulation by Selective Attention on Chromatin

• Deep learning models in genomics: are we there yet?