Dissertation Prsentation - Vaibhav

PREDICTING THE CELLULAR
LOCALIZATION SITES OF
PROTEINS USING ARTIFICIAL
NEURAL NETWORKS
Submitted by:
Vaibhav Dhattarwal
08211018
Supervisor and Guide:
Dr Durga Toshniwal

Organization of Presentation
• Introduction
• Problem Statement
• Background
• Stages Proposed
• Algorithm Implementation
• Results & Discussion
• Conclusion & Future Work
• References

Introduction
• If one is able to deduce or figure out the sub cellular
location of a protein, we can interpret its function, its
part in healthy processes and also in commencement
of disease, and it’s probable usage as a drug target.
• The sub cellular location of a protein can provide
valuable information about the role it has in the
cellular dynamics.
• The intention is to understand their basic or specific
function regards to the life of the cell.

Introduction: Protein Structure

Problem Statement
• “Prediction of Cellular Localization sites of proteins
using artificial neural networks”
• This report aims to combine the simulated artificial
neural networks and the field of bioinformatics to
predict the location of protein in a yeast genome.
• I have introduced a new sub cellular prediction
method based on a back propagation algorithm
implemented artificial neural network.

Background
• Neural Network Definition
• Neural Network Applications
• Neural Network Categorization
• Types of Neural Network
• Perceptrons and Learning Algorithm
• Classification for Yeast protein data set

Background: Neural Network Definition
• A neural network is a system which consists of many
simple processing elements that are operating in parallel
and their function ascertained by network structure,
strengths or weights of connections and the computation
done at those computing elements/nodes.
• A neural network is a massively parallel distributed
processor what holds a strong inherent ability to store
large amount of experimental knowledge. It has two
features:
▫ Knowledge is acquired through a learning procedure.
▫ Interneuron connection strengths or weights are used to
store this knowledge.

• Computer scientists can find out properties of non-symbolic
information processing by using neural networks, they can
also find out more about learning systems in general.
• Statisticians might be able to use neural networks as flexible
and nonlinear regression, and classification models.
• Neural Networks can be used by engineers for signal
processing and automatic control.
• Cognitive scientists deploy neural networks to describe
models of thinking and consciousness, which is basically brain
function.
• Neurophysiologists use neural networks to describe and
research medium-level brain function.
Background: Neural Network Applications

Background: Neural Network
Categorization

• Supervised Learning based
▫ Feed Forward Topology based
▫ Feed Back Topology based
▫ Competitive Learning based
• Unsupervised Learning based
▫ Competitive Learning based
▫ Dimension Reduction Process
▫ Auto Associative Memory
Background: Types of Neural Network

Background: Perceptrons and Learning
Algorithm

Background: Yeast Protein Data Set
• erl : It is representative of the lumen in the endoplasmic reticulum in the cell. This attribute
tells whether an HDEL pattern as n signal for retention is present or not.
• vac : This attribute give an indication of the content of amino acids in vacuolar and
extracellular proteins after performing a discriminant analysis.
• mit : This attribute gives the composition of N terminal region, which has twenty residue, of
mitochondrial as well as non-mitochondrial protein after performing a discriminant analysis.
• nuc : This feature tell us about nuclear localization patterns as to whether they are present or
not. It also holds some information about the frequency of basic residues.
• pox : This attribute provides the composition of the sequence of protein after discriminant
analysis on them. Not only this, it also indicates the presence of a short sequence motif.
• mcg : This is a parameter used in a signal sequence detection method known as McGeoch.
However in this case we are using a modified version of it.
• gvh : This attribute represents a weight matrix based procedure and is used to detect signal
sequences which are cleavable.
• alm : This final feature helps us by performing identification on the entire sequence for
membrane spanning regions.

Background: Classification Tree

Proposed Stages Of Work Done
• Stage one: Simulating the network
• Stage two: Implementing the algorithm
• Stage three: Training the Network
• Stage four: Obtaining results and comparing
performance

Stage One: Simulating the network

Stage Two: Implementing the algorithm

Stage Three: Training the network
• The localization site is represented by the class as output.
Here are the various classes:
▫ CYT (cytoskeletal)
▫ NUC (nuclear)
▫ MIT (mitochondrial)
▫ ME3 (membrane protein, no N-terminal signal)
▫ ME2 (membrane protein, uncleaved signal)
▫ ME1 (membrane protein, cleaved signal)
▫ EXC (extracellular)
▫ VAC (vacuolar)
▫ POX (peroxisomal)
▫ ERL (endoplasmic reticulum lumen)

Stage Four: Obtaining Results &
Comparing Performance
• The yeast data set class statistics are mapped as
output.
• The attributes of the data set are mapped to reflect the
variation of output.
• Varying the number of nodes in the hidden layer is
used to evaluate performance.
• Parameters for comparing performance are:
▫ Accuracy on test set.
▫ Ratio of correctly classified in the training set

Algorithm Implementation
• Sigmoid Function & Its derivative
• Pseudo Code for a single network layer
• Pseudo Code for all network layers
• Pseudo Code for training patterns
• Pseudo Code for minimizing error

Sigmoid Function & Its derivative

Pseudo Code for a single network layer
• InputLayer2[j] = Wt[0][j]
• for all elements in Layer One [ NumUnitLayer1 ]
• do
▫ Add to InputLayer2[j] the sum over the product
OutputLayer1[i] * Wt[i][j]
• end for
• Compute the sigmoid to get activation output

Pseudo Code for all network layers
• for all elements in hidden layer [ NumUnitHidden ] // computes Hidden Layer PE outputs //
• do
▫ InputHidden[j] = WtInput/Hidden[0][j]
▫ for all elements in input layer [ NumUnitInput ]
▫ do
 Add to InputHidden[j] the sum over OutputInput[i] * WtInput/Hidden [i][j]
▫ end for
• Compute sigmoid for output
• end for
• for all elements in output layer [ NumUnitOutput ] // computes Output Layer PE outputs //
• do
▫ InputOutput[k] = WtHidden/Output[0][k]
▫ for all elements in hidden layer [ NumUnitHidden ]
▫ do
 Add to InputOutput [k] sum over OutputHidden[j] * WtHidden/Output [j][k]
▫ end for
• Compute sigmoid for output
• end for

Pseudo Code for training patterns
• Er = 0.0 ;
• for all patterns in the training set
• do // computes for all training patterns(E) //
▫ do
 InputHidden[E][j] = WtInput/Hidden[0][j]
 for all elements in input layer [ NumUnitInput ]
 do
 Add to InputHidden[E] [j] the sum over OutputInput[E] [i] * WtInput/Hidden
[i][j];
 end for
▫ Compute sigmoid for output
▫ end for

Pseudo Code for training patterns
▫ for all elements in output layer [ NumUnitOutput ]
▫ do
 InputOutput[E] [k] = WtHidden/Output[0][k]
 for all elements in hidden layer [ NumUnitHidden ]
 do
 Add to InputOutput [E] [k] sum over OutputHidden[E] [j] *
WtHidden/Output [j][k]
 end for
▫ Compute sigmoid for output
▫ Add to Er the sum over the product (1/2) * (Final[E][k] -
Output[E][k]) * (Final[E][k] - Output[E][k])
▫ end for
• end for

Pseudo Code for minimizing error
• for all elements in hidden layer [ NumUnitHidden ]
• do // This loop updates the weight input to hidden
//
▫ Add to ΔWih [0][j] the sum of: product β * ΔH [j] to the
product: α * ΔWih [0][j]
▫ Add to WtInput/Hidden [0][j] the change ΔWih [0][j]
▫ for all elements in input layer [ NumUnitInput ]
▫ do
 Add to ΔWih [i][j] the sum of product β * InputHidden [p][i] * ΔH
[j] to the product: α * ΔWih [i][j]
▫ Add to WtInput/Hidden [i][j] the change ΔWih [i][j]
▫ end for
• end for

Pseudo Code for minimizing error
• for all elements in output layer [ NumUnitOutput ]
• do // This loop updates the weight hidden to
output //
▫ Add to ΔWho [0][k] the sum of: product β * ΔOutput[k] to
the product: α * ΔWho [0][k]
▫ Add to WtHidden/Output [0][k] the change ΔWho [0][k]
▫ do
 Add to ΔWho [j][k] the sum of product β * OutputHidden [p][j] *
ΔOutput [k] to the product: α *ΔWho [j][k]
 Add to WtHidden/Output [j][k] the change ΔWho [j][k]
▫ end for
• end for

Results & Discussion
• Yeast Data Set Class Statistics
• Yeast Data Set Attributes
• Comparison of Accuracies of various algorithms
• Variation of success rate with number of iterations
• Variation of success rate with number of nodes in
hidden layer
• Variation of accuracy in training with the criteria in
testing

Results: Yeast Data Set class statistics

Results: Yeast Data Set Attributes

Results: Comparison of Accuracies of
various algorithms

Results: Variation of Success Rate with no
of iterations

Results: Variation of Success Rate with no
of nodes in Hidden Layer

Results: Variation of Accuracy in Training
with Criteria in Testing

Conclusion
• The classes CYT, NUC and MIT have the largest
number of instances.
• Interesting observations are that the value of erl and
pox are almost constant throughout the entirety of the
data set whereas the rest of the attributes show
constant variation.
• The algorithm is able to achieve slightly higher
accuracy than the rest of the algorithms.

Conclusion
• Another thing of note is to see that considerable
success is achieved in the yeast data set which we
chose to implement with accuracy leading up to 61%
• After about 100 iterations the success rate remains
constant more or less.
• The success rate reaches a constant value after about
75 elements in the layer.
• The Accuracy rises till we reach the limit to which we
can set the success rate.

Future Work
• Since the prediction of proteins’ cellular localization
sites is a typical classification problem, many other
techniques such as probability model, Bayesian
network, K-nearest neighbours etc, can be compared
with our technique.
• Thus, an aspect of future work is to examine the
performance of these techniques on this particular
problem.

Key References
• [1]. "A Probablistic Classification System for Predicting the Cellular Localization Sites of
Proteins", Paul Horton & Kenta Nakai, Intelligent Systems in Molecular Biology, 109-115.
• [2]. "Expert Sytem for Predicting Protein Localization Sites in Gram-Negative Bacteria",
Kenta Nakai & Minoru Kanehisa, PROTEINS: Structure, Function, and Genetics 11:95-110,
1991.
• [3]. "A Knowledge Base for Predicting Protein Localization Sites in Eukaryotic Cells", Kenta
Nakai & Minoru Kanehisa, Genomics 14:897-911, 1992.
• [4]. Cairns, P. Huyck, et.al, A Comparison of Categorization Algorithms for Predicting the
Cellular Localization Sites of Proteins, IEEE Engineering in Medicine and Biology, pp.296-
300, 2001.
• [5]. Donnes, P., and Hoglund, A., Predicting protein subcellular localization: Past, present,
and future Genomics Proteomics Bioinformatics, 2:209-215, 2004.
• [6]. Feng, Z.P., An overview on predicting the subcellular location of a protein, In Silico
Biol2002.
• [7]. Reinhardt, A., and Hubbard, T., Using neural networks for prediction of the subcellular
location of proteins, Nucleic Acids Res., 26(9):2230-2236, 1998.

Dissertation Prsentation - Vaibhav

More Related Content

What's hot (19)

Similar to Dissertation Prsentation - Vaibhav (20)

More from Vaibhav Dhattarwal (6)

Dissertation Prsentation - Vaibhav