Weighted Markov Chain Based Aggregation of Bio-molecule Orderings

Area of Science:

• Bioinformatics
• Computational Biology
• Data Science

Background:

• Rank aggregation is crucial for meta-analysis in bioinformatics, combining disparate results from various sources like gene prediction algorithms or microarray studies.
• Existing rank aggregation methods, such as Markov chains (MC) and evolutionary algorithms, have limitations, particularly in handling weighted sources.
• There is a need for a formal framework for weighted Markov chains to improve the accuracy and confidence in aggregated rankings.

Purpose of the Study:

• To propose a novel rank aggregation method that incorporates expert-defined weights for different data sources.
• To adapt existing Markov chain (MC) approaches for weighted aggregation scenarios.
• To enhance the effectiveness of meta-analysis in bioinformatics by providing a more nuanced approach to combining ranked data.

Main Methods:

• Utilizing a modified version of the MC4 Markov chain algorithm.
• Applying a weighted analog of local Kemenization for the aggregation process.
• Incorporating expert-assigned weights to prioritize different sources of rankings.

Main Results:

• The proposed method enables weighted rank aggregation, addressing a gap in current Markov chain applications.
• The approach allows for expert-driven prioritization of data sources, leading to potentially more accurate aggregated rankings.
• Demonstrates the feasibility of combining MC with weighted Kemenization for enhanced meta-analysis.

Conclusions:

• The developed method offers a robust framework for weighted rank aggregation in bioinformatics.
• This approach improves meta-analysis by allowing the integration of expert knowledge through source weighting.
• The modified MC4 and weighted Kemenization provide a faster and more flexible alternative to evolutionary algorithms for weighted aggregation.