DL - Deep Neural Network Framework

A Deep Neural Network Framework built to solve the DREAM-ENCODE transcription factor binding site prediction challenge:

Inspired by work in the Deep Learning Specialization course by Andrew Ng

Overview

A configurable feed-forward neural network to classify genomic regions as TF bound or unbound, using data from sequence and chromatin accessibility features. Deep learning fundamentals backed by real genomics ML benchmark.

Highlights

• Configurable multi-layer neural network
• End-to-end training and evaludation pipeline

Tech Stack

Python, NumPy, Pandas, Keras

Data

The ENCODE-DREAM challenge priveds the ChIP-seq binding labels plus DNase-seq and sequence features.