Protein Secondary Structure Prediction and framework for neural network based web service

1
Protein Secondary Structure Predictionand
framework for neural network based web service

• 4-29-2004

2
Neural Network Web Service

• Web service application that accepts client
  request consisting of the network identifier and
  query string.
• Reads in network file in rdf format that
  specifies information about the network and any
  sub-networks including nodes, edges, weights,
  activation functions.
• Goal is to be able to use it for any type of
  neural network, although currently only limited
  functionality has been implemented because of
  time constraints.

3
Sequence Profile Web Service

• Creates a profile of amino acid occurrences in
  homologous sequences to be used as input for the
  secondary structure prediction network.
• Profiles are created from sequences obtained from
  a BLAST (Basic Local Alignment Search Tool)
  search from the NCBI (National Center for
  Biotechnology Information) servers.
• Uses the BLAST URL API to send a request to the
  NCBI servers. The file returned is then processed
  to separate the sequence data.

4
Sequence Profile Web Service

• The sequences returned from the BLAST search are
  then compared to determine the occurrences of
  amino acids at each position of the sequence.
• The web service returns profiles for each
  position in the sequence.

Example profile for one position in the sequence
5
Secondary Structure Prediction Client

• Web page that allows the user to input a sequence
  of amino acids and request a prediction of the
  secondary structure.
• Connects to the neural network web service
  application.
• Interface will need some later refinements.
• Related information based on the query can be
  added later.

6
Training of neural network for secondary
structure prediction

• Windows application for training and testing
  neural network.
• Still working on this part!

7
Issues and Further Work

• Neural network web service still has a bugs that
  need to be worked out. All web application are
  still in need of some error handling routines.
• Had some technical difficulties in trying to use
  the Drive API with the web service and had to use
  a different method for reading rdf files. Need to
  re-examine this later.
• Performance of the web services are still an
  issue (very slow). Some refinements may be
  necessary to improve performance. Some
  performance issues are related to the NCBI
  servers.
• Need to finish windows application and train the
  secondary structure network.

8
Issues and Further Work

• Expand upon design of current rdf model of neural
  network for more complex networks (i.e. recurrent
  networks). Implement methods in web service to
  handle more neural network designs.
• Expand to include more bioinformatics services on
  the client web site.

9
Demo

• Demonstration of client web page
• http//www.hattem.us/BioWeb/bioweb.aspx

Test Sequence
MTMDKSELVQKAKLAEQAERYDDMAAAMKAVTEQGHELSNEERNLLSVAY
KNVVGARRSSWRVISSIEQKTERNEKKQQMGKEYREKIEAELQDICNDVL
ELLDKYLIPNATQPESKVFYLKMKGDYFRYLSEVASGDNKQTTVSNSQQA
YQEAFEISKKEMQPTHPIRLGLALNFSVFYYEILNSPEKACSLAKTAFDE
AIAELDTLNEESYKDSTLIMQLLRDNLTLWTSENQGDEGDAGEGEN