Protein Secondary Structure Prediction with Neural Networks: A Tutorial
Protein Secondary Structure Prediction with Neural Networks:
A Tutorial

Copyright (c) 1998-99, Adrian Shepherd, UCL

LAST UPDATE: 14/10/99

This tutorial is periodically updated by
Adrian Shepherd
( a.shepherd@biochem.ucl.ac.uk)

This is an expanded version of a talk given at
the inaugural meeting of the

Computational Molecular Biology &
Bioinformatics Discussion Group,
Joint Research School in Biomolecular Sciences,
21st March 1997

Contents
Introduction
  1. What are neural networks?
  2. Why is protein secondary structure prediction important?
  3. Why use neural networks to predict protein structures?
  4. Why not use neural networks for protein structure prediction?

Feed-Forward Networks

  1. Introduction to feed-forward nets
  2. The feed-forward architecture
  3. Training a feed-forward net

Secondary Structure Prediction with Feed-Forward Nets: the Basics

  1. The basic approach
  2. Measuring performance
  3. Looking inside the black box
  4. Drawbacks with the basic approach

Secondary Structure Prediction with Feed-Forward Nets: Beyond the Basics

  1. Adding biological information
  2. Changing the training set
  3. Post processing and filtering
  4. Changing the architecture
  5. Conclusions

Bibliography