Currently, many feature selection documents use this document based on the intuitive understanding of common information, that is, positive value indicates the existence of phenotype, increasing the degree of dependency between features, evidence of the interaction between features. negative values indicate that the existence of phenotype increases the redundancy between features. Zero represents that features are independent of each other, or, the gain and redundancy between features are mutually offset. Therefore, it is an innovative task to gain an in-depth understanding of common information in theory.
The first task in the identification phase is to use the common information to measure the interaction between SNP in the Candidate SNP set in the exhaustive search. When the co-information of all SNP and phenotype in the SNP combination is correct, it indicates that the phenotype increases the dependency between the SNP and is evidence of the interaction between the SNP combination and the phenotype. Therefore, this SNP combination is retained and further evaluated using the replacement test. When the co-information of all SNP and phenotype in the SNP combination is not correct, there is a redundant relationship between the SNP combination and the phenotype. Therefore, this SNP combination is not considered later.
The total effect of a snp on a table, including not only the effect (main effect) on the table ), this also includes the impact of this SNP on its own portion (marginal effect) in the phenotype effect through its interaction with other SNP)
Mi & Ci