GAVis: a Tool for Visualization and Control of Genetic Algorithms for -omic Data Analysis

Area of Science:

• Bioinformatics
• Computational Biology
• Machine Learning

Background:

• Genetic algorithms are computationally intensive for omics data analysis.
• Support Vector Machines (SVMs) with leave-one-out validation require significant processing time per generation.
• Parallel processing with isolated populations ('islands') can enhance diversity and computational mapping.

Purpose of the Study:

• To develop a visualization and steering application (GAVis) for parallel genetic algorithms.
• To improve the understanding and control of genetic algorithm convergence in long-term computations.
• To address the computational challenges in marker identification using omics data.

Main Methods:

• Development of the GAVis application for real-time monitoring and intervention.
• Implementation of a parallel genetic algorithm framework with "islands" for population diversity.
• Runtime steering mechanisms including inter-island migration and adaptive mutation rate adjustments.

Main Results:

• GAVis provides visualization of genetic algorithm behavior during parallel execution.
• Runtime interventions, such as migration and mutation rate adjustments, can accelerate convergence.
• The application facilitates management of computationally demanding marker identification tasks in omics data.

Conclusions:

• GAVis is an effective tool for managing and optimizing parallel genetic algorithms.
• Runtime steering strategies are crucial for efficient convergence in complex computational tasks.
• The developed methods offer a pathway to overcome computational bottlenecks in bioinformatics and machine learning applications.