Scribd
Upload
10 views

Genetic Algorithms (GA) : Group 4

Using genetic algorithms to analyze gene expression data can help: - Identify similarities and differences between cancer and normal cells by looking for consensus gene sequences and evolutionary links between proteins - Classify cancer types and provide clinical diagnoses more quickly by finding the minimal set of discriminating genes However, challenges include ensuring accuracy with limited data, managing large datasets, and requiring significant computing power for analysis. Ongoing work seeks algorithms that optimize accuracy, efficiency and generalization across applications.

Uploaded by

inflamous
Copyright
© Attribution Non-Commercial (BY-NC)
Available Formats
Download as PPT, PDF, TXT or read online on Scribd
10 views

Genetic Algorithms (GA) : Group 4

Using genetic algorithms to analyze gene expression data can help: - Identify similarities and differences between cancer and normal cells by looking for consensus gene sequences and evolutionary links between proteins - Classify cancer types and provide clinical diagnoses more quickly by finding the minimal set of discriminating genes However, challenges include ensuring accuracy with limited data, managing large datasets, and requiring significant computing power for analysis. Ongoing work seeks algorithms that optimize accuracy, efficiency and generalization across applications.

Uploaded by

inflamous
Copyright
© Attribution Non-Commercial (BY-NC)
Available Formats
Download as PPT, PDF, TXT or read online on Scribd
You are on page 1/ 20

Genetic Algorithms (GA)

Group 4
What is a Genetic Algorithm?
• It is a randomized search
and optimization technique
How are Genetic Algorithms
used?
How are Genetic Algorithms
used?
• Find genes by looking for consensus sequences
• Estimate which genes or segments are used
given a DNA fragment or RNA primer
• Find similar proteins between species to find
evolutionary links
• Find binding sites between protein sequences
Purpose of Genetic Algorithms

• Simplify sequences
into statistical
measures

• Identify trends
...more specifically

• Identify similarities between cancer cells


• Identify differences between cancer cells
and normal cells
In order to do all of
this, we need...

• Experimental data
• Microarrays
• Databases
Microarrays
• To classify test data
▫ knowing which genes you want
• Quickly provide a clinical diagnosis for cancer
• The algorithm finds the minimal set of genes to
test
• Too few genes limits the accuracy,
• Too many genes creates too much noise
Microarrays
• Discriminant analysis:
▫ method of exploring gene-expression microarray
data.
▫ requires gene selection
▫ easier to handle the large data volume,
▫ aids understanding diseases and functions of
certain genes.
Algorithms
• Pattern classification
▫ practical in medical diagnosis
• Neural networks
▫ fine tuning solutions
• Provide a compact gene set
• Understand the functions of particular genes and plan the
diagnosis process. 
• Genetic Algorithms provide
▫ more flexibility
▫ but more time consuming
Gene selection methods:
• A good feature selection method:
– low generalization error
– low computational cost
– small number of evaluations
• Feature selection method’s components
• search procedure
• evaluation function
• stopping criterion
• validation procedure
Genetic algorithm (GA)
• Pros:
▫ commonly used
▫ high-quality solution
▫ good time frame
• Cons:
▫ long time to locate the local optimum
▫ may not find the optimum with precision
Feature selection methods
• Marginal filters
– individual feature ranking methods
– the most efficient
– The selected feature subsets are sub-optimal
• Wrapper method
– a classifier is the evaluation criterion
– wrapper and embedded methods are often closely
related to each other.
– guaranteed to discover the optimal feature subset
with respect to the evaluation criterion
Issues

•Accuracy

•Data Management

•Computing Power
Accuracy
• How is a genetic algorithm assessed for accuracy?

• Main purpose is to organize data for interpretation


• what if this organization isn't accurate?

• Find which algorithms give the more optimal results


• while still keeping same function

• An algorithm's generalization
(ability to serve multiple functions)is one measure
Data Management?

• How to store data in databases?

• How to organize all this information for


intuitive and accessible use?

• New techniques in storing


-> international collaboration with data!

• Ex: NCBI, Biomart


Computing Power
• Processing power is required for research to be
conducted

• Larger computer power = larger number of


scripts and programs handling data set with
many iterations that can be run
Which algorithms came out
on top?

⁃ kNN algorithm performs better than


more sophisticated methods in various
situations

⁃ Evolutionary approach can improve


accuracy and reduce computational
overhead
Next Developments: Medical Diagnosis

You might also like