Multiobjective Rank Aggregation for Gene Prioritization

Rank aggregation involves consolidating multiple individual preference rankings of items to generate a consensus ranking. Typically framed as an optimization problem, it seeks to minimize the average distance between the aggregated ranking and the input rank lists to achieve consensus. However, it is important to recognize that minimizing the average distance may not guarantee an unbiased aggregation. This is because the aggregated ranking could closely resemble one input ranking while being distant from another, leading to a biased overall result with reduced average distance. This chapter presents a technique that treats the rank aggregation problem as a multiobjective optimization challenge. In this approach, the objective is not only to minimize the average distance but also to simultaneously minimize the standard deviation of the distance values, aiming to avoid bias toward any specific input ranking. To achieve this, a multiobjective particle swarm optimization (PSO) is adopted to develop a rank aggregation algorithm. The performance of this technique is showcased using artificial datasets and its ability is demonstrated for gene ranking from microarray gene expression data, comparing its effectiveness with state-of-the-art techniques to establish its superiority.