Open AccessDisease model identification methods based on maximum test and performance analysis
Author(s) -
Yaling Yin
Publication year - 2020
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1656/1/012020
Subject(s) - computer science , statistic , computation , statistics , entropy (arrow of time) , test statistic , statistical hypothesis testing , artificial intelligence , data mining , machine learning , mathematics , algorithm , physics , quantum mechanics
Combinatorial explosion and computational burden are always the challenges for genome-wide association study. In order to reduce the computation cost, many multi-stage methods were put forward to identify the disease models. However, one-way and two-way disease models always can be detected to leave out some SNPs for non-significance. And these SNPs are combined with other SNPs to get higher disease models. In this paper, three test statistics, Max Gtest, Max Entropy Difference and Max Relative Entropy, had been presented for the first stage to detection disease models with main effect and without main effect. Five testing methods were used for examining multiply simulation datasets and real dataset. Results were revealed that Max Entropy Difference test is the best method of recognition in five filtering methods with main-effect and max-statistic test is just right method to identify model without main-effect. Results also were showed that five statistics can get interest power for two-ways on simulation datasets and real dataset. We believe that these statistics can find strong and weak SNPs for next step in computationally and statistically.