The Shapley and Banzhaf values in microarray games
Introduction
It is well known that several diseases have a genetic origin. Recently, a new tool, called microarray technology, has been implemented to generate a lot of information on gene expression of human beings. These data can in principle be used to identify genes which are, for instance, responsible of a particular disease. Many models for data analysis have been presented in the literature for inferring, from a matrix of gene expression data, the role of genes, their interactions and their behaviors when changes in condition of the biological system occur (see for instance the book [12]). An alternative method for gene expression analysis based on coalitional games has been proposed in [8]. The main advantage of this approach is the possibility to compute a numerical index, called a relevance index, which represents the relevance of each gene under a certain condition (e.g., a tumor) taking into account the expression behaviors of the other genes. In [8], the authors use the Shapley value with the aim of ranking the genes according to their relevance and provide an axiomatic characterization of the Shapley value on the class of the so-called microarray games.
In literature, other approaches using game theory for gene expression analysis have been introduced. In [7], a method based on the framework of minimum cost spanning trees (MCST) has been introduced to represent the interactions between all possible pairs of genes and extended to implement the notion of association for coalitions of genes. In [4], classification games with genes in the role of players have been studied to analyze the power of groups of genes to classify samples into the right classes (for instance, the class of normal tissues or the class of tumor tissues). A biological validation for the use of the Shapley value ([13]; see also [9]) of microarray games as a relevance index for genes is presented in [1], where a set of genes involved in the pathogenesis of neuroblastic tumors has been selected according to the Shapley value of a microarray game. The problem to compare the relevance of genes under two different conditions has been recently studied in [10] where a statistical method called comparative analysis of Shapley (shortly, CASh) value is applied to real gene expression data concerning the gene expression in children differentially exposed to air pollution.
In this paper we consider another important one-point solution in game theory, namely the Banzhaf value [3]. In particular, taking also inspiration from [6], we offer a pool of properties characterizing it, and we give an alternative characterization for the Shapley value, always on the class of the microarray games. Finally, we compare the results given by the two indices on a set of data relative to patients affected by colon tumor.
Section snippets
Preliminary notations
Let us introduce some basic game theoretical notations. A coalitional game or characteristic-form game is a pair , where denotes the finite set of players and the characteristic function, with . If the set of players is fixed, we identify a coalitional game with the corresponding characteristic function . A group of players is called a coalition and is called the value of this coalition. A coalitional game such that is called a -game.
Axiomatic characterization for microarray games
In [8] it was proved that the Shapley value is the only one-point solution, on the class of microarray games, fulfilling a pool of reasonable properties that we shall describe later. In this paper we are interested in giving another pool of properties characterizing the Banzhaf value. In doing this, we also find another alternative characterization of the Shapley value. To start with, we introduce a new definition, motivated by an analogous one given in [5], for general cooperative games. Definition 1 Let
Some thoughts on Banzhaf versus Shapley
The two relevance indices we deal with in this paper are suitable to rank genes potentially responsible of a genetic disease. In general, they will give different ranking. How can we interpret this fact? In this section we briefly comment on this.
It is clear that the differences in the two indices arise from the differences of their behavior with respect to the unanimity games. So, what is the basic difference among them, when dealing with this type of games? Of course, they do assign zero to
Colon data analysis
Moretti et al. [8] introduced a preliminary application of the Shapley value for a microarray game defined on a tumor/normal data-set published in [2]5 containing expression levels of a set of 2000 genes measured using Affymatrix oligonucleotide microarrays for a set of tumor samples and a set of 22 normal samples, in total 62 samples from colon tissues. In that application, after the preprocessing stage performed by the
Acknowledgments
The authors are grateful to two anonymous referees for their helpful comments. Stefano Moretti gratefully acknowledges the financial support of the EU project NewGeneris, European Union 6th FP (FOOD-CT-2005-016320).
References (13)
- et al.
A game theoretical approach to the classification problem in gene expression data analysis
Computers & Mathematics with Applications
(2008) - et al.
Identification of low intratumoral gene expression heterogeneity in neuroblastic tumors by wide-genome expression analysis and game theory
Cancer
(2008) - et al.
Broad patterns of gene expression revealed by clustering analysis of tumor and normal colon tissue probed by oligonucleotide arrays
Proceedings of the National Academy of Sciences of the United States of America
(1999) Weighted voting doesn’t work: a game theoretic approach
Rutgers Law Review
(1965)- Kalai E, Samet D. Weighted Shapley values. In: Roth A, editor. The Shapley value, essays in honor of Lloyd S. Shapley....
- et al.
Shapley–Shubik and Banzhaf indices revisited
Mathematics of Operations Research
(2001)
Cited by (39)
The Choquet integral-based Shapley function for n-person cooperative games with probabilistic hesitant fuzzy coalitions
2023, Expert Systems with ApplicationsMulti-robot task allocation clustering based on game theory
2023, Robotics and Autonomous SystemsCitation Excerpt :To relieve this issue, we propose a clustering approach to group sets of robots and tasks, which become the players of a game. The proposed method is based on the ordering of players introduced by the Shapley value [45], the best-known solution concept in cooperative game theory, with applications in water systems [46], biology [47], finance [48], or power networks [49], among others. Certainly, the Shapley value has been applied to multi-agent systems to form coalitions due to its relationship with the marginal contribution of each agent [50].
Coalitional games for networked controllers with constraints on semivalues: A randomized design approach
2022, Journal of the Franklin InstituteAssessing systematic sampling in estimating the Banzhaf–Owen value
2020, Operations Research LettersCitation Excerpt :For specific TU-games, particular expressions of the coalitional values have been obtained for their exact computation, reducing the arisen difficulty in large-scale problems. [14] and [25] determine the exact expressions of the Shapley and the Owen values for the airport games and [11] obtain the exact expression of the Shapley value for the maintenance games. [15] characterize the Shapley value and the Banzhaf value on the class of the microarray games as an application in genetic.
Efficient design of fault detection architectures for power networks by using game theory
2020, IFAC-PapersOnLine
- 1
The research of this author was partially supported by Ministero dell’Istruzione, dell’Università e della Ricerca Scientifica (COFIN 2005).