Hidetoshi Matsui

Variable selection for functional regression models via the L1 regularization

In regression analysis, L1 regularizations such as the lasso or the SCAD provide sparse solutions, which leads to variable selection. We consider the variable selection problem where variables are given as functional forms, using L1 regularization. In order to select functional variables each of which is controlled by multiple parameters, we treat parameters as grouped parameters and then apply the group SCAD. A crucial issue in the regularization method is the choice of regularization parameters. We derive a model selection criterion for evaluating the model estimated by the regularization method via the group SCAD penalty. Results of simulation and real data analysis show the effectiveness of the proposed modeling strategy.

Functional Logistic Discrimination Via Regularized Basis Expansions

We introduce a functional logistic discrimination based on basis expansions with the help of regularization, which classifies functional data into several distinct groups. A crucial issue in model building process is the choice of regularization parameters. Choosing these parameters can be viewed as a model selection and evaluation problem. We derive a Bayesian model selection criterion for evaluating models estimated by the method of regularization in the context of functional logistic discrimination. Monte Carlo experiments are conducted to examine the efficiency of the proposed functional discrimination procedure. We also apply our procedure to the analysis of yeast cell cycle microarray data. The results show that our modeling procedure provides useful tools for classifying functions or curves.

Variable and boundary selection for functional data via multiclass logistic regression modeling

Penalties with an l1 norm provide solutions in which some coefficients are exactly zero and can be used for selecting variables in regression settings. When applied to the logistic regression model, they also can be used to select variables which affect classification. We focus on the form of l1 penalties in logistic regression models for functional data, in particular, their use in classifying functions into three or more groups while simultaneously selecting variables or classification boundaries. We provide penalties that appropriately select the variables in functional multiclass logistic regression models. Analysis of simulation and real data show that the form of the penalty should be selected in accordance with the purpose of the analysis.

MicroRNAs Induce Epigenetic Reprogramming and Suppress Malignant Phenotypes of Human Colon Cancer Cells

Although cancer is a genetic disease, epigenetic alterations are involved in its initiation and progression. Previous studies have shown that reprogramming of colon cancer cells using Oct3/4, Sox2, Klf4, and cMyc reduces cancer malignancy. Therefore, cancer reprogramming may be a useful treatment for chemo- or radiotherapy-resistant cancer cells. It was also reported that the introduction of endogenous small-sized, non-coding ribonucleotides such as microRNA (miR) 302s and miR-369-3p or -5p resulted in the induction of cellular reprogramming. miRs are smaller than the genes of transcription factors, making them possibly suitable for use in clinical strategies. Therefore, we reprogrammed colon cancer cells using miR-302s and miR-369-3p or -5p. This resulted in inhibition of cell proliferation and invasion and the stimulation of the mesenchymal-to-epithelial transition phenotype in colon cancer cells. Importantly, the introduction of the ribonucleotides resulted in epigenetic reprogramming of DNA demethylation and histone modification events. Furthermore, in vivo administration of the ribonucleotides in mice elicited the induction of cancer cell apoptosis, which involves the mitochondrial Bcl2 protein family. The present study shows that the introduction of miR-302s and miR-369s could induce cellular reprogramming and modulate malignant phenotypes of human colorectal cancer, suggesting that the appropriate delivery of functional small-sized ribonucleotides may open a new avenue for therapy against human malignant tumors.

Multivariate Regression Modeling for Functional Data

We propose functional multivariate regression modeling, using Gaussian basis functions along with the technique of regularization. In order to evaluate the model estimated by the regularization method, we derive model selection criteria from information-theoretic and Bayesian viewpoints. Monte Carlo simulations are conducted to investigate the efficiency of the proposed model. We also apply our modeling strategy to the analysis of spectrometric data.

Metabolic Adaptation to Nutritional Stress in Human Colorectal Cancer

Tumor cells respond to their microenvironment, which can include hypoxia and malnutrition, and adapt their metabolism to survive and grow. Some oncogenes are associated with cancer metabolism via regulation of the related enzymes or transporters. However, the importance of metabolism and precise metabolic effects of oncogenes in colorectal cancer remain unclear. We found that colorectal cancer cells survived under the condition of glucose depletion, and their resistance to such conditions depended on genomic alterations rather than on KRAS mutation alone. Metabolomic analysis demonstrated that those cells maintained tricarboxylic acid cycle activity and ATP production under such conditions. Furthermore, we identified pivotal roles of GLUD1 and SLC25A13 in nutritional stress. GLUD1 and SLC25A13 were associated with tumor aggressiveness and poorer prognosis of colorectal cancer. In conclusion, GLUD1 and SLC25A13 may serve as new targets in treating refractory colorectal cancer which survive in malnutritional microenvironments.

Mathematical analysis predicts imbalanced IDH1/2 expression associates with 2-HG-inactivating β-oxygenation pathway in colorectal cancer

Bioinformatics and computational modeling offer innovative approaches to investigate cancer metabolism and predict the secondary and tertiary cellular responses. Dysregulation of metabolism has also been implicated in the pathophysiology of cancer. A significant proportion of patients with glioblastoma and hematological malignancies harbor the mutated forms of the oxidative phosphorylation (OxPhos) enzymes, isocitrate dehydrogenase (IDH) 1 or 2. The mutated forms of IDH1 and IDH2 produce an oncogenic metabolite, D-2-hydroxyglutarate (D2HG). A recent study of breast cancer patients showed that D2HG can also be produced in the absence of mutated IDH, through an alternative route involving over-activated MYC signaling. We developed a novel methodology to computationally analyze gene expression in colorectal cancer (CRC), and identified novel sets of genes that are associated with patient survival. The study of OxPhos-related genes revealed that an imbalance between the expression of IDH1 and IDH2, defined as overexpression of one isoform in relation to the other, was associated with worse prognosis in CRC patients. This effect was further accentuated by reduced expression of the β-oxygenation enzyme, 3-D-hydroxyacyl-CoA dehydratase (HCDH) 4, which has been reported to contribute to metabolism of intracellular D2HG. The present computational analysis revealed a novel and potential mechanism of CRC development, through over-production of D2HG when there is an imbalance between IDH1 and IDH2 expression, resulting in decreased clearance of D2HG when the β-oxidization pathway is diminished. Additional validation analysis with another gene expression dataset resulted in IDH1/2 imbalanced expression with a shorter DFS compared with balanced expression. Altogether, these findings provide a strong rationale for studying this mechanism further in order to discover novel therapeutic targets for the treatment of CRC.

Gene set differential analysis of time course expression profiles via sparse estimation in functional logistic model with application to time-dependent biomarker detection

High-throughput time course expression profiles have been available in the last decade due to developments in measurement techniques and devices. Functional data analysis, which treats smoothed curves instead of originally observed discrete data, is effective for the time course expression profiles in terms of dimension reduction, robustness, and applicability to data measured at small and irregularly spaced time points. However, the statistical method of differential analysis for time course expression profiles has not been well established. We propose a functional logistic model based on elastic net regularization (F-Logistic) in order to identify the genes with dynamic alterations in case/control study. We employ a mixed model as a smoothing method to obtain functional data; then F-Logistic is applied to time course profiles measured at small and irregularly spaced time points. We evaluate the performance of F-Logistic in comparison with another functional data approach, i.e. functional ANOVA test (F-ANOVA), by applying the methods to real and synthetic time course data sets. The real data sets consist of the time course gene expression profiles for long-term effects of recombinant interferon β on disease progression in multiple sclerosis. F-Logistic distinguishes dynamic alterations, which cannot be found by competitive approaches such as F-ANOVA, in case/control study based on time course expression profiles. F-Logistic is effective for time-dependent biomarker detection, diagnosis, and therapy.

Embryonic MicroRNA-369 Controls Metabolic Splicing Factors and Urges Cellular Reprograming

Noncoding microRNAs inhibit translation and lower the transcript stability of coding mRNA, however miR-369 s, in aberrant silencing genomic regions, stabilizes target proteins under cellular stress. We found that in vitro differentiation of embryonic stem cells led to chromatin methylation of histone H3K4 at the miR-369 region on chromosome 12qF in mice, which is expressed in embryonic cells and is critical for pluripotency. Proteomic analyses revealed that miR-369 stabilized translation of pyruvate kinase (Pkm2) splicing factors such as HNRNPA2B1. Overexpression of miR-369 stimulated Pkm2 splicing and enhanced induction of cellular reprogramming by induced pluripotent stem cell factors, whereas miR-369 knockdown resulted in suppression. Furthermore, immunoprecipitation analysis showed that the Argonaute complex contained the fragile X mental retardation-related protein 1 and HNRNPA2B1 in a miR-369-depedent manner. Our findings demonstrate a unique role of the embryonic miR-369-HNRNPA2B1 axis in controlling metabolic enzyme function, and suggest a novel pathway linking epigenetic, transcriptional, and metabolic control in cell reprogramming.

Multiclass Functional Discriminant Analysis and Its Application to Gesture Recognition

We consider applying a functional logistic discriminant procedure to the analysis of handwritten character data. Time-course trajectories corresponding to the X and Y coordinate values of handwritten characters written in the air with one finger are converted into a functional data set via regularized basis expansion. We then apply functional logistic modeling to classify the functions into several classes. In order to select the values of adjusted parameters involved in the functional logistic model, we derive a model selection criterion for evaluating models estimated by the method of regularization. Results indicate the effectiveness of our modeling strategy in terms of prediction accuracy.