Chapter 1 Introduction to Deep Learning

1.1 What is Deep Learning?

Deep Learning (DL) is a subfield of machine learning based on artificial neural networks. The “deep” refers to the number of layers through which the data is transformed. While a standard neural network might have 2-3 layers, a deep network can have dozens or hundreds.

1.1.1 Key Concepts

  • Neuron: The basic computational unit. It takes inputs, multiplies them by weights, adds a bias, and passes the result through an activation function.
  • Layer: A collection of neurons.
  • Input Layer: The layer that receives the raw data.
  • Hidden Layer(s): The layers between input and output where the “learning” happens.
  • Output Layer: The final layer that produces the prediction.
  • Activation Function: A function that determines the output of a neuron (e.g., ReLU, Sigmoid, Softmax).

1.2 Why Deep Learning in Bioinformatics?

Bioinformatics data is often high-dimensional and complex, making it a perfect candidate for DL.

Some examples:

  • Genomics: Predicting the functional effect of non-coding variants, chromatin accessibility, and transcription factor binding.
  • Proteomics: Predicting protein structure (e.g., AlphaFold), function, and interactions.
  • Medical Imaging: Classifying tumors from histopathology images or MRI scans.
  • Drug Discovery: Predicting molecular properties and drug-target interactions.

1.3 A Simple Biological Example: Transcription Factor Binding Prediction

Imagine we have DNA sequences of length 100bp. Our goal is to predict whether a specific transcription factor (TF) binds to a given sequence.

  • Input: "ATCGATCGAT..." (100 characters)
  • Output: 1 (binds) or 0 (does not bind)

We need to:

  1. Convert the sequence into a numerical format (more on this in Chapter 3).
  2. Design a neural network that can learn the binding motif and its context.

In the following chapters, we will learn how to build such a model step-by-step.