In this paper, we study a stepwise feature selection algorithm for a high-order interaction model and we propose a new statistical inference for selected high-order interaction features. Feature selection and statistical inference for high-order interaction features are challenging tasks because the possible number of those interactions is extremely large. Our main contribution is to extend recently developed selective inference framework to high-order interaction model by developing a pruning technique for searching over tree which represents high-order interaction features. We demonstrate the effectiveness of the proposed approach by applying it to several synthetic problems and an HIV drug resistance prediction problem.
Metabolic engineering strategies enabling the production of specific target metabolites by host strains can be identified in silico through the use of metabolic network analysis such as flux balance analysis. This type of metabolic redesign is based on the computation of reactions that should be deleted from the original network representing the metabolism of the host strain to enable the production of the target metabolites while still ensuring its growth (the concept of growth coupling). In this context, it is important to develop algorithms that enable this growth-coupled reaction deletions identification for any metabolic network topologies and any potential target metabolites. A recent method that ensures the target metabolite production even when the cell growth is not maximized (strong coupling) has been shown to be able to identify such computational redesign for nearly all metabolites included in the genome-scale metabolic models of Escherichia coli and Saccharomyces cerevisiae when cultivated under aerobic conditions. However, this approach enables the computational redesign of S. cerevisiae for only 3.9% of all metabolites if under anaerobic conditions. Therefore, it is necessary to develop algorithms able to perform for various culture conditions. The author developed an algorithm, CubeProd, that could calculate the reaction deletions that achieve the coupling of growth and production under the condition that the cell growth is maximized (weak coupling) for 91.3% metabolites in genome-scale models of S. cerevisiae under anaerobic conditions. In CubeProd, the solution space was divided into small sub-spaces by the constraints on cell growth, target production, and the absolute sum of fluxes, and the reaction deletion strategies that achieve weak coupling were efficiently determined. While the weak coupling-based methods assume the cell growth maximization, the strong coupling-based methods do not assume it. Computational experiments showed that the proposed algorithm is efficient also for aerobic conditions and E. coli. The developed software, CubeProd, implemented in MATLAB, and the obtained reaction deletion strategies are freely available on https://sunflower.kuicr.kyoto-u.ac.jp/~tamura/software.
Neurological problems can manifest in body movements. Pronation and supination of the forearms are used to evaluate the performance of diadochokinesis. Diagnosis in the clinic is subjective and nonquantitative. Introducing a sinusoidal visual stimulus, a novel measuring and evaluating method for human motor control function in a motor synchronization process has been proposed. The Non-Smoothness Measure (NSM) and movement speed measure (P3r) are confirmed in this study as promising evaluation parameters based on experiments on age groups.