A heuristic for gene selection and visual prediction of sample type

Area of Science:

• Bioinformatics
• Computational Biology
• Data Visualization

Background:

• Accurate gene selection is crucial for understanding biological states.
• Traditional statistical and machine learning methods can lack intuitive interpretability.
• Visual analytics offer potential to enhance insights from complex biological data.

Purpose of the Study:

• To introduce a heuristic gene selection method integrated with RadViz visualization.
• To enable visual prediction of tissue samples in normal and disease states.
• To enhance the interpretability of gene selection and predictive modeling.

Main Methods:

• A heuristic gene selection approach.
• Integration with RadViz for visual data representation.
• Development of a visual analytics framework for the entire predictive modeling pipeline.

Main Results:

• The integrated method allows intuitive visualization of gene selection, model training, evaluation, and prediction.
• Visual analytics provide enhanced insights compared to classical methods.
• The case study demonstrates cost-effectiveness and competitive performance against existing techniques.

Conclusions:

• The proposed visual analytics method offers an intuitive and effective approach for gene selection and tissue sample prediction.
• This integrated visual approach enhances the interpretability and utility of machine learning in bioinformatics.
• The method is both cost-effective and performs competitively with established techniques.