Advertisement
Browse Subject Areas
?

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

Application of a Combination of a Knowledge-Based Algorithm and 2-Stage Screening to Hypothesis-Free Genomic Data on Irinotecan-Treated Patients for Identification of a Candidate Single Nucleotide Polymorphism Related to an Adverse Effect

  • Hiro Takahashi ,

    hiro.takahashi@chiba-u.jp

    Affiliations Graduate School of Horticulture, Chiba University, Matsudo, Chiba, Japan, Plant Biology Research Center, Chubu University, Kasugai, Aichi, Japan, Division of Genetics, National Cancer Center Research Institute, Tokyo, Japan

  • Kimie Sai,

    Affiliation Division of Medicinal Safety Science, National Institute of Health Sciences, Tokyo, Japan

  • Yoshiro Saito,

    Affiliation Division of Medicinal Safety Science, National Institute of Health Sciences, Tokyo, Japan

  • Nahoko Kaniwa,

    Affiliation Division of Medicinal Safety Science, National Institute of Health Sciences, Tokyo, Japan

  • Yasuhiro Matsumura,

    Affiliation Division of Developmental Therapeutics, Research Center for Innovative Oncology, National Cancer Center Hospital East, Kashiwa, Chiba, Japan

  • Tetsuya Hamaguchi,

    Affiliation Gastrointestinal Medical Oncology Division, National Cancer Center Hospital, Tokyo, Japan

  • Yasuhiro Shimada,

    Affiliation Gastrointestinal Medical Oncology Division, National Cancer Center Hospital, Tokyo, Japan

  • Atsushi Ohtsu,

    Affiliation Department of Gastrointestinal Oncology, National Cancer Center Hospital East, Kashiwa, Chiba, Japan

  • Takayuki Yoshino,

    Affiliation Department of Gastrointestinal Oncology, National Cancer Center Hospital East, Kashiwa, Chiba, Japan

  • Toshihiko Doi,

    Affiliation Department of Gastrointestinal Oncology, National Cancer Center Hospital East, Kashiwa, Chiba, Japan

  • Haruhiro Okuda,

    Affiliation Division of Medicinal Safety Science, National Institute of Health Sciences, Tokyo, Japan

  • Risa Ichinohe,

    Affiliations Division of Genetics, National Cancer Center Research Institute, Tokyo, Japan, Faculty of Horticulture, Chiba University, Matsudo, Chiba, Japan

  • Anna Takahashi,

    Affiliation Plant Biology Research Center, Chubu University, Kasugai, Aichi, Japan

  • Ayano Doi,

    Affiliations Division of Genetics, National Cancer Center Research Institute, Tokyo, Japan, Faculty of Horticulture, Chiba University, Matsudo, Chiba, Japan

  • Yoko Odaka,

    Affiliation Division of Genetics, National Cancer Center Research Institute, Tokyo, Japan

  • Misuzu Okuyama,

    Affiliation Division of Genetics, National Cancer Center Research Institute, Tokyo, Japan

  • Nagahiro Saijo,

    Current address: Japanese Society of Medical Oncology, Tokyo, Japan

    Affiliation National Cancer Center Hospital East, Kashiwa, Chiba, Japan

  • Jun-ichi Sawada,

    Current address: Pharmaceutical and Medical Devices Agency, Tokyo, Japan

    Affiliation Division of Functional Biochemistry and Genomics, National Institute of Health Sciences, Tokyo, Japan

  • Hiromi Sakamoto,

    Affiliation Division of Genetics, National Cancer Center Research Institute, Tokyo, Japan

  •  [ ... ],
  • Teruhiko Yoshida

    Affiliation Division of Genetics, National Cancer Center Research Institute, Tokyo, Japan

  • [ view all ]
  • [ view less ]

Application of a Combination of a Knowledge-Based Algorithm and 2-Stage Screening to Hypothesis-Free Genomic Data on Irinotecan-Treated Patients for Identification of a Candidate Single Nucleotide Polymorphism Related to an Adverse Effect

  • Hiro Takahashi, 
  • Kimie Sai, 
  • Yoshiro Saito, 
  • Nahoko Kaniwa, 
  • Yasuhiro Matsumura, 
  • Tetsuya Hamaguchi, 
  • Yasuhiro Shimada, 
  • Atsushi Ohtsu, 
  • Takayuki Yoshino, 
  • Toshihiko Doi
PLOS
x

Abstract

Interindividual variation in a drug response among patients is known to cause serious problems in medicine. Genomic information has been proposed as the basis for “personalized” health care. The genome-wide association study (GWAS) is a powerful technique for examining single nucleotide polymorphisms (SNPs) and their relationship with drug response variation; however, when using only GWAS, it often happens that no useful SNPs are identified due to multiple testing problems. Therefore, in a previous study, we proposed a combined method consisting of a knowledge-based algorithm, 2 stages of screening, and a permutation test for identifying SNPs. In the present study, we applied this method to a pharmacogenomics study where 109,365 SNPs were genotyped using Illumina Human-1 BeadChip in 168 cancer patients treated with irinotecan chemotherapy. We identified the SNP rs9351963 in potassium voltage-gated channel subfamily KQT member 5 (KCNQ5) as a candidate factor related to incidence of irinotecan-induced diarrhea. The p value for rs9351963 was 3.31×10−5 in Fisher's exact test and 0.0289 in the permutation test (when multiple testing problems were corrected). Additionally, rs9351963 was clearly superior to the clinical parameters and the model involving rs9351963 showed sensitivity of 77.8% and specificity of 57.6% in the evaluation by means of logistic regression. Recent studies showed that KCNQ4 and KCNQ5 genes encode members of the M channel expressed in gastrointestinal smooth muscle and suggested that these genes are associated with irritable bowel syndrome and similar peristalsis diseases. These results suggest that rs9351963 in KCNQ5 is a possible predictive factor of incidence of diarrhea in cancer patients treated with irinotecan chemotherapy and for selecting chemotherapy regimens, such as irinotecan alone or a combination of irinotecan with a KCNQ5 opener. Nonetheless, clinical importance of rs9351963 should be further elucidated.

Introduction

Genomic information has been proposed to be utilized as the basis for “personalized” health care. Interindividual variation in a drug response among patients has been well documented to cause serious problems in pharmacotherapy. This variation may be due to multiple factors such as disease phenotypes, genetic and clinical parameters (or environmental factors), and variability in the drug target or allergic response; all of these factors may affect both main and side effects [1], [2]. Although some biomarkers [3][9] have been proposed, it is still difficult to determine which group of patients will respond positively, which patients are nonresponders, and which may experience adverse reactions in cases where patients are administered the same medication dose. For effectiveness of personalized medicine in cancer chemotherapy, it is critically important to observe interindividual differences in a drug response and the role of genetic polymorphisms relevant to the drug metabolic pathways and drug response biology in pharmacogenomics [10].

Irinotecan (CPT-11), an anticancer prodrug, is widely used for the treatment of a broad range of carcinomas, such as colorectal, lung, ovarian, and cervical cancers. Unexpected severe diarrhea and neutropenia are prominent adverse effects of irinotecan treatment. The active metabolite SN-38 (7-ethyl-10-hydroxycamptothecin), a topoisomerase I inhibitor, is generated via hydrolysis of the parent compound by carboxylesterases [11], and is subsequently glucuronidated by uridine diphosphate glucuronosyltransferases (UGTs), such as UGT1A1, UGT1A7, or UGT1A9, to form an inactive metabolite, SN-38 glucuronide (SN-38G) [12][14]. Irinotecan is also inactivated by CYP3A4 to produce 7-ethyl-10- [4-N-(5-aminopentanoic acid)-1-piperidino] carbonyloxycamptothecin (APC; a major CYP3A4 product) and 7-ethyl-10-(4-amino-1-piperidino) carbonyloxycamptothecin (NPC; a minor product) [15], [16]. Irinotecan and its metabolites are excreted into the bile and urine via the action of ATP-binding cassette (ABC) transporters, such as P-glycoprotein (P-gp/ABCB1), multiple resistance-associated protein 2 (MRP2/ABCC2), and breast cancer resistance protein (BCRP/ABCG2) [17]. Transport of SN-38 from the plasma into the liver is mediated by the organic anion transporting polypeptide C (OATP-C/SLCO1B1) [18]. Most of the previous pharmacogenetic studies of irinotecan have been focused on UGT1A1 polymorphisms and have shown clinical relevance of UGT1A1*28, a repeat polymorphism in the TATA box [-54_-39A(TA)6TAA>A(TA)7TAA or -40_-39ins TA], to severe adverse effects [3], [19], [20]. Based on these findings, in 2005, the Food and Drug Administration (FDA) of the United States approved an amendment for the formulation called Camptosar (irinotecan-HCl) (NDA 20-571/S-024/S-027/S-028) and for clinical use of a genetic diagnostic kit for the *28 allele. In parallel with this advance in the USA, clinical relevance to severe neutropenia of UGT1A1*6 [211G>A(G71R)], another low-activity allele detected specifically in East Asians, as well as *28, was demonstrated in several studies on Asian patients [5], [21][23]. Accordingly, in June 2008, the Ministry of Health, Labor, and Welfare of Japan approved changes to irinotecan formulations (Campto and Topotecin) by adding a warning about the risk of severe adverse effects in patients either homozygous or compound-heterozygous for UGT1A1*28 and *6 (*28/*28, *6/*6, *28/*6) and also approved clinical use of a diagnostic kit for UGT1A1*28 and *6. Severe adverse effects, however, are reported in patients without the genetic variations *6/*6, *28/*28, and *28/*6; therefore, several clinical studies have suggested that polymorphisms of the drug transporter genes, such as ABCB1, ABCC2, ABCG2, and SLCO1B1, might affect irinotecan pharmacokinetics (PK)/pharmacodynamics (PD) in Caucasian and Asian patients [22], [24][35], as shown in Fig. 1. Nonetheless, the almost all reported results deal with PK in patients and neutropenia induced by irinotecan as an adverse reaction not but with diarrhea. Therefore, other factors responsible for other irinotecan adverse effects, such as diarrhea should be identified.

Diarrhea induced by irinotecan is classified into early- and delayed-onset diarrhea, occurring within 24 hr or ≤24 hr after irinotecan administration, respectively [36]. Irinotecan induces early-onset diarrhea as one of adverse cholinergic effects (acetylcholinelike effects) by inhibiting acetylcholinesterase (AChE) and binding to muscarinic acetylcholine receptors (mAChR) [37], [38]. These inhibitory actions are induced by irinotecan, which has an amino group at the C-10 position [37]. Other than that, irinotecan induces delayed-onset diarrhea via rapid deconjugation of SN38G and adsorption of released free SN-38 by β-glucuronidase of intestinal flora [39][41], as shown in Fig. 1. In the present study, we focused on polymorphisms of genes with transporter activity to identify predictive factors of diarrhea induced by irinotecan because there are many genes related to transporter activity in both pathways.

A genome-wide association study (GWAS), also known as a whole-genome association study (WGA study, or WGAS), is an examination of many common genetic variants in different individuals to determine whether a particular variant is associated with a trait. GWAS using hypothesis-free genomic data is a powerful technique for identifying interindividual variation among patients. On the other hand, multiple testing problems are a limitation of this approach. To address this issue, we recently proposed a combined method consisting of a knowledge-based algorithm, 2 stages of screening, and a permutation test for identifying single nucleotide polymorphisms (SNPs) [7]. In general, the objective of a statistical or bioinformatic analysis is the enrichment of important information in a large dataset [42][47]. The use of a knowledge-based algorithm is not a novel concept, but is both practical and useful [48][66]. In the previous study, we found that rs2293347 in the gene of human epidermal growth factor receptor (EGFR) is a candidate SNP related to the chemotherapeutic response; we achieved this result by applying our combined method to gastric cancer patients who were treated with fluoropyrimidine [7]. However, our combined method was applied to only 1 dataset. Therefore, the usability of our combined method as a novel approach was still unclear.

We used the combined method in an actual genome-wide pharmacogenomics study of antitumor drugs, particularly irinotecan. We found that rs9351963 in the gene of potassium voltage-gated channel subfamily KQT member 5 (KCNQ5) is a candidate SNP related to the adverse response. Rs9351963 may be a potential predictive factor of incidence of diarrhea in cancer patients treated with the cancer prodrug irinotecan.

Materials and Methods

Ethics statement

The study was conducted according to the principles expressed in the Declaration of Helsinki, and the ethics committees of the National Cancer Center and National Institute of Health Sciences, Japan, approved the study protocol. All patients provided written informed consent to participate.

Preparation of hypothesis-free genomic data for cancer patients treated with irinotecan

This study was performed within the framework of the Millennium Genome Project in Japan, and 4 antitumor drugs were chosen as project targets: gemcitabine, paclitaxel, fluoropyrimidine, and irinotecan. These drugs (alone or in some combination) were administered to approximately 1,000 cancer patients at the National Cancer Center, Japan. Additionally, approximately 1,000 DNA samples were extracted from peripheral blood mononuclear cells and 109,365 SNPs were genotyped using the Illumina Human-1 BeadChip. In this study, we focused on pharmacogenomic properties of irinotecan. Participants included 177 Japanese irinotecan-naïve cancer patients (56 cancer patients treated with irinotecan monotherapy and 121 cancer patients treated with irinotecan combination therapy) at the National Cancer Center Hospital and National Cancer Center Hospital East. A summary of the characteristics of the 176 patients is listed in Table S1. We excluded 1 patient who refused grading of adverse reactions. Furthermore, we excluded 8 patients who did not have genotyping data. Therefore, we analyzed the remaining 168 patients (53 cancer patients treated with irinotecan monotherapy and 115 cancer patients treated with irinotecan combination therapy) in the present study. We defined the 53 patients treated with irinotecan monotherapy as the first dataset and the 168 patients treated with irinotecan chemotherapy (consisting of irinotecan monotherapy and combination therapy) as the second dataset for 2 stages of screening.

Monitoring and adverse effects

A complete medical history and data on physical examination were recorded before the irinotecan therapy. Complete blood cell counts with differentials and platelet counts, as well as blood biochemical variables, were measured once a week during the first 2 months of irinotecan treatment. Adverse events were graded according to the National Cancer Institute - Common Toxicity Criteria (NCI-CTC Version 2.0). Only the highest grade of adverse events was recorded during the first 2 months of irinotecan treatment for each patient and adverse event.

Patient characteristics and clinical parameters

A summary of the patients' characteristics in the two datasets for diarrhea is shown in Table 1. The association of genetic or clinical parameters with incidence of grade ≥2 diarrhea was examined on the basis of Spearman's rank correlation coefficient. “UGT1A1*6 or *28” is an effective genetic predictive factor of irinotecan-induced neutropenia and pharmacokinetics in cancer patients [5]. This factor was constructed from 2 polymorphisms: UGT1A1*6 and *28.

thumbnail
Table 1. Irinotecan-treated cancer patients with SNP information, genetic factor, and clinical parameters for incidence of diarrhea.

https://doi.org/10.1371/journal.pone.0105160.t001

Fisher's exact test

This statistical test is usually used to determine nonrandom associations between 2 categorical variables [67]. Fisher's exact test is similar to the chi-squared test. If a sample size is large, then the chi-squared test is suitable. Nevertheless, significance values from the chi-squared test are only approximated. Fisher's exact test is used in to analyze contingency tables when the sample sizes are small [67]. We used Fisher's exact test in the present study. The odds ratio (OR) is defined as a×d/(b×c), where a is the number of patients that had adverse events with a minor allele, b is the number of patients that did not have adverse events with a minor allele, c is the number of patients that had adverse events with a major allele, and d is the number of patients that did not have adverse events with a major allele. The null hypothesis for Fisher's exact test is OR = 1.

The permutation test

The permutation test theory evolved from the works of Fisher and Pitman in the 1930s [68]. In this study, p values of multiple-comparison analyses were adjusted by applying the permutation test to 2 stages of screening. The case–control (or phenotype) labels were randomly shuffled for the 2 screening stages, and p values were calculated using Fisher's exact test. The lowest p value was selected for the randomized data. This procedure was repeated 100,000 times. Exact p values for the permutation test were calculated based on the distribution of the lowest p values.

Multiple testing correction

The Bonferroni correction is a method used to address the problem of multiple comparisons (also known as the multiple testing problem). It is considered the simplest and most conservative method for control of the family-wise error rate (FWER). In addition, false discovery rate (FDR) controlling procedures, such as the Benjamini-Hochberg (BH) method [69], are more powerful (i.e., less conservative) than the FWER procedures, such as the Bonferroni correction, at the cost of increasing the likelihood of false positives within the rejected hypothesis. In the present study, the BH method was used to calculate the q value. The q value is defined as an FDR analog of the p value.

The Akaike information criterion (AIC)

The AIC is a measure of the relative goodness of fit of a statistical model [70]. A smaller AIC indicates a better fit when comparing fitted objects. The AIC is defined according to the formula -2× (log likelihood) + (2×npar), where npar represents the number of parameters in the fitted model, and the log-likelihood value [71] is obtained from the logistic regression model.

The receiver operating characteristic (ROC)

ROC analysis is a graphical plot that illustrates the performance of a binary classifier system as its discrimination threshold is varied. It is built by plotting sensitivity (the number of true positive results divided by the number of true positive samples) against (1 minus specificity) at various threshold settings. (Specificity is the number of true negative results divided by the number of true negative samples.) The area under the curve (AUC) of a ROC curve is an indicator of expected performance of the test. A higher AUC is more desirable, with a value of 1.00 denoting perfect performance (sensitivity and specificity are both 100%), while a value of 0.50 indicates random performance.

Gene set based on gene ontology GO terms

GO has been developed to provide scientists with a controlled terminology system for labeling gene functions in a precise, reliable, computer-readable manner. Data for annotated genes and associated GO terms were obtained from the GO website (http://www.geneontology.org). We compiled a GO term list to select polymorphisms in genes with transporter activity (GO:0005215) and related activities, as shown in Table S2. The numbers of GO terms obtained was 943. GO data were obtained on July 1, 2010.

Results

Association analysis of adverse affects and clinical parameters (or a genetic factor)

The association between clinical parameters (or a genetic factor) and incidence of grade ≥2 diarrhea was examined on the basis of Spearman's rank correlation coefficient, as shown in Table 1. This table shows that no parameter was associated with the adverse response to chemotherapy (incidence of grade ≥2 diarrhea) in the first dataset (patients treated with irinotecan monotherapy). Nonetheless, Amrubicin (p = 0.00625) was significantly associated with the response in the second dataset (patients treated with any irinotecan chemotherapy: a combination or monotherapy). Mitomycin C (MMC; p = 0.0740) was weakly associated with the response. These clinical factors should be evaluated when constructing diagnostic models involving multiple factors.

Extraction of candidate SNPs using the combined method consisting of the knowledge-based algorithm, 2 stages of screening, and the permutation test

In this study, we applied the combined method to hypothesis-free genomic data on cancer patients treated with irinotecan chemotherapy as shown in Fig. 2. Figure 2A shows an outline of the knowledge-based algorithm for identifying SNPs (KB-SNP). In the previous study, we extracted rs numbers (SNP IDs) related to cancer using a combination of National Center for Biotechnology Information (NCBI) dbSNP and NCBI PubMed [7]. In the present study, we extracted rs numbers from genes linked to specific GO terms instead of the combination of NCBI dbSNP and PubMed. In this analysis, we defined specific GO terms as the terms related to transporter activity.

thumbnail
Figure 2. An outline of chemotherapeutic response analysis in irinotecan-treated cancer patients using a combined method consisting of a knowledge-based algorithm for identifying SNPs (KB-SNP), 2 stages of screening, and the permutation test.

(A) The KB-SNP algorithm. (B) Extraction of SNPs using KB-SNP for statistical analysis. (C) Two-stage screening of irinotecan-treated cancer patients. (D) Calculation of the p value in the permutation test based on the 2 stages of screening. The second dataset (includes first dataset) was permutated. The permutated first dataset was extracted from the permutated second dataset. By using Fisher's exact test, SNPs with p<0.005 for the first dataset were selected from among 5,242 SNPs. Among the selected SNPs, those with the lowest p value in Fisher's exact test for the second dataset were selected. This procedure was repeated 100,000 times and empirical null distribution was constructed. Using the distribution, the actual p value obtained from the second stage of screening was converted to the adjusted p value (based on correction of multiple testing problems). At these screening steps, allele models were used for each SNP.

https://doi.org/10.1371/journal.pone.0105160.g002

A total of 6,506 SNPs related to transporter activity were extracted from 109,365 SNPs using KB-SNP (Fig. 2B). Furthermore, we excluded SNPs with a p value <0.2 in the Hardy-Weinberg equilibrium (HWE) or the minor allele frequency (MAF) <0.05. Then the extracted 5,242 SNPs were used in the association study.

We analyzed 53 patients treated with irinotecan monotherapy as the first dataset for first-stage screening in the association study (Fig. 2C). Each p value was calculated using Fisher's exact test for the allele model. A total of 24 SNPs with p<0.005 were extracted. In the second stage of screening, 168 patients treated with irinotecan chemotherapy (including 53 patients treated with irinotecan monotherapy) were analyzed to validate these 24 SNPs. Adjustment of a calculated p value for the second stage of screening was conducted using the permutation test for these 2 stages of screening (Fig. 2D). Only rs9351963 in KCNQ5 showed a statistically significant p value (0.0289), which was determined using the permutation test. The rs9351963 is a common variant (MAF = 0.328). Furthermore, we conducted Fisher's exact test and used the Benjamini-Hochberg method [69] to calculate p and q values for the second dataset only. Seven SNPs had a q value <1, as shown in Table 2. Six SNPs (rs11022922, rs3918305, rs3813627, rs768172, rs3813628, and rs10815019) had q = 0.802 as shown in Table 2. This result indicates that 5 out of 6 SNPs were false positive; however, we assessed performance of only rs9351963 in the process of model construction.

Comparison of models based on rs9351963 in KCNQ5

We analyzed not only an allele model but also dominant and recessive models of rs9351963 in KCNQ5 in relation to the first dataset (irinotecan monotherapy), the second dataset (any irinotecan chemotherapy), and the dataset of irinotecan combination chemotherapy (excluding irinotecan monotherapy), as shown in Figure 3. Figure 3A shows that the p value of the allele model was the lowest (p = 8.86×10−5, OR = 6.3), and the p value (p = 1.29×10−4, OR = 24) of the dominant model was lower than the p value (p = 0.0358, OR = 7.0) of the recessive model in the first dataset. In addition, Figure 3B shows that the p value of the allele model was the lowest (p = 3.31×10−5, OR = 3.1), and the p value (p = 1.28×10−4, OR = 6.7) of the recessive model was lower than the p value (p = 4.44×10−3, OR = 3.3) of the dominant model in the second dataset. Therefore, we evaluated the 3 models using the dataset of irinotecan combination chemotherapy (excluding irinotecan monotherapy; Fig. 3C). Figure 3C shows that the p value (p = 1.44×10−3, OR = 6.9) of the recessive model meant strong statistical significance and the OR was almost equal to OR ( = 7.0) in the first dataset, as shown in Figure 3A. Although ORs of the recessive models seemed to have high homogeneity among all 3 datasets, there was no statistical evidence. Therefore, the proportional odds model was used to construct multiple logistic regression models.

thumbnail
Figure 3. Contingency tables for rs9351963 in KCNQ5 for each model using each dataset.

(A) irinotecan monotherapy (first dataset), (B) any irinotecan chemotherapy (including irinotecan monotherapy; second dataset), and (C) irinotecan combination chemotherapy (excluding irinotecan monotherapy). OR: odds ratio. The p values were calculated using Fisher's exact test. CI: confidence interval.

https://doi.org/10.1371/journal.pone.0105160.g003

Selection of the model of rs9351963 in KCNQ5 and construction of multiple regression models

We compared the AICs and AUCs using the second dataset in the 8 models: NULL (without parameter), “UGT1A1*6 or *28” (an integrated predictive factor based on polymorphisms related to neutropenia), and rs9351963 (genotype of rs9351963 in KCNQ5), Amrubicin, MMC, rs9351963+Amrubicin, rs9351963+MMC, and rs9351963+Amrubicin+MMC (Fig. 4A). Figure 4A shows that performance of all models except UGT1A1 *6 or *28 is better than the performance of the NULL model. Although the Amrubicin+MMC (combination of Amrubicin and MMC) model was better than Amrubicin alone or MMC, the rs9351963 models were clearly better than the Amrubicin+MMC model, as shown in Figures 4A and 4B. Performance of rs9351963+Amrubicin and rs9351963+MMC models was better than performance of the rs9351963 model. Furthermore, performance of the rs9351963+Amrubicin+MMC model was better than that of rs9351963+Amrubicin and rs9351963+MMC models. Therefore, we selected the rs9351963+Amrubicin+MMC model as the best one on the basis of AIC. AUC, sensitivity, and specificity of this model were 0.744, 77.8%, and 57.6% in in the ROC curve, respectively, as shown in Figures 4A and 4B.

thumbnail
Figure 4. Comparison of AIC, AUC, and ROC curves for logistic regression models.

(A) Parameters of each model. (B) The ROC curve of a model consisting of rs9351963+MMC+ Amrubicin. ROC: receiver operating characteristic, AUC: area under the ROC curve, NULL indicates the model without parameters. Each genetic factor conforms to the proportional odds model, AIC: Akaike's information criterion, AUC: area under the ROC curve, Sens.: Sensitivity (%), Spec.: Specificity (%).

https://doi.org/10.1371/journal.pone.0105160.g004

Discussion

In the present study, we used 2 stages of screening: the method that is based on the concept of joint analysis. Joint analysis is more efficient than replication-based analysis [72]. The first dataset is a part of the second dataset in joint analysis (the latter includes the former). In contrast, the 2 datasets must be independent in a replication-based analysis (which we did not use here). Our 2 stages of screening derived from the joint analysis were used to increase statistical detection power. KB-SNP was performed prior to 2 stages of screening. KB-SNP reduced the number of candidate SNPs to 6,506 from 109,365. Approximately 80,000 SNPs can be extracted without knowledge-based reduction of the SNP number. Thus, statistically significant SNPs cannot be extracted from the present data. We could find the statistically significant rs9351963 in KCNQ5 by applying the combined method to hypothesis-free genomic data.

The KCNQ/K(v)7 potassium channel family consists of 5 members of neural muscarine channel (M channel; from KCNQ1 to KCNQ5) which have a distinct expression pattern and a functional role. Although KCNQ1 is prevalently expressed in the cardiac muscle, KCNQ2, KCNQ3, KCNQ4, and KCNQ5 are expressed in neural tissue [73][75]. On the other hand, a recent study revealed that KCNQ4 and KCNQ5 are the most abundantly expressed KCNQ channels in smooth muscle throughout the gastrointestinal tract [76]. Furthermore, Jepps et al. opined that drugs that selectively block KCNQ4/KCNQ5 might be promising as therapeutics for the treatment of motility disorders such as constipation associated with irritable bowel syndrome [76]. In other words, drugs that selectively open KCNQ4/KCNQ5 might be effective against diarrhea. The KCNQ family gene products assemble as homomeric or heteromeric tetramers to form functional channels that mediate the M-current [77], a current that is suppressed by mAChR activation [78], [79]. Irinotecan induces adverse cholinergic effects (acetylcholinelike actions) by inhibiting AChE and binding to mAChR [37], [38], [80]. Therefore, polymorphisms of KCNQ5 genes possibly effect incidence of diarrhea as interindividual variation in the drug response among cancer patients treated with irinotecan chemotherapy.

In the present study, only the highest grade of adverse events is recorded during the first 2 months of irinotecan treatment for each patient and each adverse effect. Therefore, incidence of grade ≥2 diarrhea possibly includes cases caused partially by enterohepatic circulation of APC and NPC, but genotype of rs9351963 in KCNQ5 correlates with the start date of treatment with antidiarrheal agents (Spearman's rank correlation coefficient ρ = −0.198, p = 0.00995). In other words, genotype of rs9351963 may correlates with the diagnosis (or presentiment) of irinotecan induced early-onset diarrhea (diagnosis is made by trained chemotherapists).

The rs9351963 A>C polymorphism is located in an intron, which does not change the amino acid sequence of the protein and may not influence the biological function of the protein itself. Nonetheless, some intronic polymorphisms are effective markers: For example, rs2237892 in intron 15 of KCNQ1 is associated with susceptibility to type 2 diabetes mellitus in Japanese individuals [81], and the CA simple sequence repeat in intron 1 (CA-SSR1) of the gene of epidermal growth factor receptor (EGFR) is associated with the clinical outcome in gefitinib-treated Japanese patients with non-small cell lung cancer [82]. Furthermore, variations related to intronic or synonymous SNPs possibly affect mRNA stability, translational kinetics, and splicing, resulting in alterations at the protein level, e.g., changes of structure or function [83][89]. Although rs9351963 does not have a known function, this SNP is a possible predictive factor of adverse effects of irinotecan-based chemotherapy and is possibly linked to some functional polymorphisms in KCNQ5. Their clinical importance needs to be further elucidated.

In the present study, we extracted rs9351963, which showed a p value (0.0289) obtained using a combination of 2 stages of screening and a permutation test from SNPs selected by KB-SNP. In the second dataset, the p value of Fisher's exact test was 3.31×10−5, and the q value was 0.173 calculated by correction of Benjamini-Hochberg method, as shown in Table 2. This value is statistically insignificant. Therefore, during the 2 stages of screening, it is statistically sufficient to extract rs9351963.

The calculation of probability of occurrence in Bernoulli trials is suitable to for estimation of validity of the repetition number in the permutation process. In this trial, occurrence probability is defined as nCk× (pB)k×(1 - pB)(nk), where k is the occurrence number, n is the repetition number, and pB represents probability. If the repetition number is 100,000 for rs9351963 (p = 0.02891 [2891/100000]) and the significance level of the test (α) is 0.05, the occurrence probability is 100000C2891× (0.05)2891× (1–0.05)(100000–2891)  = 4.89×10−241. In statistics, the 99% (or 95%) confidence interval should be considered. The significance level of α = 0.05 does not exist in the 99% confidence interval of the p value for rs9351963, because the occurrence probability 4.89×10−241 is clearly lower than 0.01. Similarly, if the repetition number is 10,000, the occurrence probability is 3.41×10−26. This way, the occurrence probability is sufficiently low for 10,000 permutations. Nevertheless, we conducted 100,000 permutations to estimate p values more accurately for the permutation test.

Using our combined method involving KB-SNP, we identified rs9351963 as a potential predictive factor of diarrhea in cancer patients treated with irinotecan chemotherapy; however, the comprehensiveness of KB-SNP was limited. Therefore, statistical information regarding the adverse effects of cancer patients treated with irinotecan chemotherapy is shown in Table S3 for incidence of diarrhea (p<0.05) and in Table S4 for incidence of neutropenia (p<0.05). The relevant data are also provided on the website Genome Medicine Database of Japan (GeMDBJ) [90] (https://gemdbj.nibio.go.jp/). These data will be useful for replication studies or meta-analyses in the future.

In conclusion, in the present study, we applied the combined method to hypothesis-free genomic data on cancer patients treated with irinotecan chemotherapy. By means of this method, rs9351963 in KCNQ5 was extracted as a candidate SNP related to the incidence of diarrhea. For example, the association of rs9351963 with irinotecan-related diarrhea (OR of 3.14) showed a p value of 3.31×10−5 in Fisher's exact test (allele model). Even if this p value were adjusted by means of the permutation test for the effects of multiple testing problems, the adjusted p value would still indicate statistical significance (adjusted p value of 0.0289<0.05). Additionally, we evaluated the performance of rs9351963 using multiple regression models. rs9351963 was clearly superior to clinical parameters (or environmental factors) and showed a sensitivity of 77.8% and specificity of 57.6% in the multiple regression model, including rs9351963. Recent studies showed that the KCNQ4 and KCNQ5 genes encode components of the M channel expressed in gastrointestinal smooth muscles and suggested that these genes are associated with irritable bowel syndrome and similar peristalsis diseases. These results suggest that rs9351963 may be a predictive factor of diarrhea in cancer patients treated with irinotecan chemotherapy. This SNP may also be useful for selection of chemotherapy regimens, such as irinotecan monotherapy or a combination of irinotecan chemotherapy with KCNQ5 opener. Furthermore, the result of the present analysis supports usability of our combined method.

Supporting Information

Table S1.

Irinotecan-treated cancer patients, genetic factor, and clinical parameters for incidence of diarrhea and neutropenia.UGT1A1*6 or *28” is a genetic factor constructed from 2 polymorphisms (UGT1A1*6 and *28); “2” indicates *6/*6, *28/*28 or *6/*28, “1” indicates *6 or *28, and “0” indicates “other than 2 and 1.” Area: body surface area (m2), PS: performance status, Cr: grade of creatinine, Hg: grade of hemoglobin, Alb: grade of albumin, ALP: grade of alkaline phosphatase, and GOT: grade of glutamic oxaloacetic transaminase. Each laboratory test value (Alb, Hg, GOT, ALP, and Cr) was recorded before the irinotecan therapy. For each type of clinical tests the grade and aberrant values were defined according to the National Cancer Institute - Common Toxicity Criteria (NCI-CTC, Version 2.0). Cmax/dose: SN38 Cmax/dose [10−3×m2/L]. AUC: area under the concentration-time curve. AUC ratio: Ratio of AUCSN38/AUCCPT-11. 5-FU: 5-fluorouracil, CDDP: cisplatin, MMC: mitomycin C, VP16: etoposide. * and † indicate p<0.05 and 0.05≤p<0.10, respectively. For each concomitant drug, 0 means “not administered,” 1 indicates administered.

https://doi.org/10.1371/journal.pone.0105160.s001

(XLS)

Table S2.

GO term list for transporter activity and the related functions.

https://doi.org/10.1371/journal.pone.0105160.s002

(XLS)

Table S3.

Statistical information on the chemotherapeutic response (incidence of grade ≥2 diarrhea) of irinotecan-treated cancer patients (p<0.05). RS number: reference SNP identification number in dbSNP; p values were calculated using Fisher's exact test and q values were calculated using the Benjamini-Hochberg (BH) method from p values.

https://doi.org/10.1371/journal.pone.0105160.s003

(XLS)

Table S4.

Statistical information on the chemotherapeutic response (incidence of grade ≥3 neutropenia) of irinotecan-treated cancer patients (p<0.05). RS number: reference SNP identification number in dbSNP; p values were calculated using Fisher's exact test and q values were calculated using the Benjamini-Hochberg (BH) method from p values.

https://doi.org/10.1371/journal.pone.0105160.s004

(XLS)

Acknowledgments

We thank Ms. Sumiko Ohnami for help with SNP genotyping.

Author Contributions

Conceived and designed the experiments: HT Y. Saito NS JS HS T. Yoshida. Performed the experiments: HT KS NK HO YO MO. Analyzed the data: HT RI AT AD. Contributed reagents/materials/analysis tools: HT YM TH Y. Shimada AO T. Yoshino TD. Contributed to the writing of the manuscript: HT.

References

  1. 1. Evans WE, McLeod HL (2003) Pharmacogenomics–drug disposition, drug targets, and side effects. N Engl J Med 348: 538–549.
  2. 2. Ingelman-Sundberg M (2008) Pharmacogenomic biomarkers for prediction of severe adverse drug reactions. N Engl J Med 358: 637–639.
  3. 3. Ando Y, Saka H, Ando M, Sawa T, Muro K, et al. (2000) Polymorphisms of UDP-glucuronosyltransferase gene and irinotecan toxicity: a pharmacogenetic analysis. Cancer Res 60: 6921–6926.
  4. 4. Ma F, Sun T, Shi Y, Yu D, Tan W, et al. (2009) Polymorphisms of EGFR predict clinical outcome in advanced non-small-cell lung cancer patients treated with Gefitinib. Lung Cancer 66: 114–119.
  5. 5. Minami H, Sai K, Saeki M, Saito Y, Ozawa S, et al. (2007) Irinotecan pharmacokinetics/pharmacodynamics and UGT1A genetic polymorphisms in Japanese: roles of UGT1A1*6 and *28. Pharmacogenet Genomics 17: 497–504.
  6. 6. Sato Y, Laird NM, Nagashima K, Kato R, Hamano T, et al. (2009) A new statistical screening approach for finding pharmacokinetics-related genes in genome-wide studies. Pharmacogenomics J 9: 137–146.
  7. 7. Takahashi H, Kaniwa N, Saito Y, Sai K, Hamaguchi T, et al. (2013) Identification of a candidate single-nucleotide polymorphism related to chemotherapeutic response through a combination of knowledge-based algorithm and hypothesis-free genomic data. J Biosci Bioeng 116: 768–773.
  8. 8. van Kuilenburg AB, Muller EW, Haasjes J, Meinsma R, Zoetekouw L, et al. (2001) Lethal outcome of a patient with a complete dihydropyrimidine dehydrogenase (DPD) deficiency after administration of 5-fluorouracil: frequency of the common IVS14+1G>A mutation causing DPD deficiency. Clin Cancer Res 7: 1149–1153.
  9. 9. Raida M, Schwabe W, Hausler P, Van Kuilenburg AB, Van Gennip AH, et al. (2001) Prevalence of a common point mutation in the dihydropyrimidine dehydrogenase (DPD) gene within the 5′-splice donor site of intron 14 in patients with severe 5-fluorouracil (5-FU)- related toxicity compared with controls. Clin Cancer Res 7: 2832–2839.
  10. 10. Efferth T, Volm M (2005) Pharmacogenetics for individualized cancer chemotherapy. Pharmacol Ther 107: 155–176.
  11. 11. Slatter JG, Su P, Sams JP, Schaaf LJ, Wienkers LC (1997) Bioactivation of the anticancer agent CPT-11 to SN-38 by human hepatic microsomal carboxylesterases and the in vitro assessment of potential drug interactions. Drug Metab Dispos 25: 1157–1164.
  12. 12. Iyer L, King CD, Whitington PF, Green MD, Roy SK, et al. (1998) Genetic predisposition to the metabolism of irinotecan (CPT-11). Role of uridine diphosphate glucuronosyltransferase isoform 1A1 in the glucuronidation of its active metabolite (SN-38) in human liver microsomes. J Clin Invest 101: 847–854.
  13. 13. Ciotti M, Basu N, Brangi M, Owens IS (1999) Glucuronidation of 7-ethyl-10-hydroxycamptothecin (SN-38) by the human UDP-glucuronosyltransferases encoded at the UGT1 locus. Biochem Biophys Res Commun 260: 199–202.
  14. 14. Gagne JF, Montminy V, Belanger P, Journault K, Gaucher G, et al. (2002) Common human UGT1A polymorphisms and the altered metabolism of irinotecan active metabolite 7-ethyl-10-hydroxycamptothecin (SN-38). Mol Pharmacol 62: 608–617.
  15. 15. Haaz MC, Rivory L, Riche C, Vernillet L, Robert J (1998) Metabolism of irinotecan (CPT-11) by human hepatic microsomes: participation of cytochrome P-450 3A and drug interactions. Cancer Res 58: 468–472.
  16. 16. Sai K, Saito Y, Fukushima-Uesaka H, Kurose K, Kaniwa N, et al. (2008) Impact of CYP3A4 haplotypes on irinotecan pharmacokinetics in Japanese cancer patients. Cancer Chemother Pharmacol 62: 529–537.
  17. 17. Sparreboom A, Danesi R, Ando Y, Chan J, Figg WD (2003) Pharmacogenomics of ABC transporters and its role in cancer chemotherapy. Drug Resist Updat 6: 71–84.
  18. 18. Nozawa T, Minami H, Sugiura S, Tsuji A, Tamai I (2005) Role of organic anion transporter OATP1B1 (OATP-C) in hepatic uptake of irinotecan and its active metabolite, 7-ethyl-10-hydroxycamptothecin: in vitro evidence and effect of single nucleotide polymorphisms. Drug Metab Dispos 33: 434–439.
  19. 19. Iyer L, Das S, Janisch L, Wen M, Ramirez J, et al. (2002) UGT1A1*28 polymorphism as a determinant of irinotecan disposition and toxicity. Pharmacogenomics J 2: 43–47.
  20. 20. Innocenti F, Undevia SD, Iyer L, Chen PX, Das S, et al. (2004) Genetic variants in the UDP-glucuronosyltransferase 1A1 gene predict the risk of severe neutropenia of irinotecan. J Clin Oncol 22: 1382–1388.
  21. 21. Han JY, Lim HS, Shin ES, Yoo YK, Park YH, et al. (2006) Comprehensive analysis of UGT1A polymorphisms predictive for pharmacokinetics and treatment outcome in patients with non-small-cell lung cancer treated with irinotecan and cisplatin. J Clin Oncol 24: 2237–2244.
  22. 22. Jada SR, Lim R, Wong CI, Shu X, Lee SC, et al. (2007) Role of UGT1A1*6, UGT1A1*28 and ABCG2 c.421C>A polymorphisms in irinotecan-induced neutropenia in Asian cancer patients. Cancer Sci 98: 1461–1467.
  23. 23. Sai K, Saito Y, Sakamoto H, Shirao K, Kurose K, et al. (2008) Importance of UDP-glucuronosyltransferase 1A1*6 for irinotecan toxicities in Japanese cancer patients. Cancer Lett 261: 165–171.
  24. 24. Sai K, Saito Y, Maekawa K, Kim SR, Kaniwa N, et al. (2010) Additive effects of drug transporter genetic polymorphisms on irinotecan pharmacokinetics/pharmacodynamics in Japanese cancer patients. Cancer Chemother Pharmacol 66: 95–105.
  25. 25. Mathijssen RH, Marsh S, Karlsson MO, Xie R, Baker SD, et al. (2003) Irinotecan pathway genotype analysis to predict pharmacokinetics. Clin Cancer Res 9: 3246–3253.
  26. 26. Sai K, Kaniwa N, Itoda M, Saito Y, Hasegawa R, et al. (2003) Haplotype analysis of ABCB1/MDR1 blocks in a Japanese population reveals genotype-dependent renal clearance of irinotecan. Pharmacogenetics 13: 741–757.
  27. 27. Zhou Q, Sparreboom A, Tan EH, Cheung YB, Lee A, et al. (2005) Pharmacogenetic profiling across the irinotecan pathway in Asian patients with cancer. Br J Clin Pharmacol 59: 415–424.
  28. 28. de Jong FA, Marsh S, Mathijssen RH, King C, Verweij J, et al. (2004) ABCG2 pharmacogenetics: ethnic differences in allele frequency and assessment of influence on irinotecan disposition. Clin Cancer Res 10: 5889–5894.
  29. 29. de Jong FA, Scott-Horton TJ, Kroetz DL, McLeod HL, Friberg LE, et al. (2007) Irinotecan-induced diarrhea: functional significance of the polymorphic ABCC2 transporter protein. Clin Pharmacol Ther 81: 42–49.
  30. 30. Xiang X, Jada SR, Li HH, Fan L, Tham LS, et al. (2006) Pharmacogenetics of SLCO1B1 gene and the impact of *1b and *15 haplotypes on irinotecan disposition in Asian cancer patients. Pharmacogenet Genomics 16: 683–691.
  31. 31. Takane H, Miyata M, Burioka N, Kurai J, Fukuoka Y, et al. (2007) Severe toxicities after irinotecan-based chemotherapy in a patient with lung cancer: a homozygote for the SLCO1B1*15 allele. Ther Drug Monit 29: 666–668.
  32. 32. Han JY, Lim HS, Shin ES, Yoo YK, Park YH, et al. (2008) Influence of the organic anion-transporting polypeptide 1B1 (OATP1B1) polymorphisms on irinotecan-pharmacokinetics and clinical outcome of patients with advanced non-small cell lung cancer. Lung Cancer 59: 69–75.
  33. 33. Han JY, Lim HS, Park YH, Lee SY, Lee JS (2009) Integrated pharmacogenetic prediction of irinotecan pharmacokinetics and toxicity in patients with advanced non-small cell lung cancer. Lung Cancer 63: 115–120.
  34. 34. Michael M, Thompson M, Hicks RJ, Mitchell PL, Ellis A, et al. (2006) Relationship of hepatic functional imaging to irinotecan pharmacokinetics and genetic parameters of drug elimination. J Clin Oncol 24: 4228–4235.
  35. 35. Sai K, Itoda M, Saito Y, Kurose K, Katori N, et al. (2006) Genetic variations and haplotype structures of the ABCB1 gene in a Japanese population: an expanded haplotype block covering the distal promoter region, and associated ethnic differences. Ann Hum Genet 70: 605–622.
  36. 36. Yang X, Hu Z, Chan SY, Chan E, Goh BC, et al. (2005) Novel agents that potentially inhibit irinotecan-induced diarrhea. Curr Med Chem 12: 1343–1358.
  37. 37. Kawato Y, Sekiguchi M, Akahane K, Tsutomi Y, Hirota Y, et al. (1993) Inhibitory activity of camptothecin derivatives against acetylcholinesterase in dogs and their binding activity to acetylcholine receptors in rats. J Pharm Pharmacol 45: 444–448.
  38. 38. Hyatt JL, Tsurkan L, Morton CL, Yoon KJ, Harel M, et al. (2005) Inhibition of acetylcholinesterase by the anticancer prodrug CPT-11. Chem Biol Interact 157–158: 247–252.
  39. 39. Takakura A, Kurita A, Asahara T, Yokoba M, Yamamoto M, et al. (2012) Rapid deconjugation of SN-38 glucuronide and adsorption of released free SN-38 by intestinal microorganisms in rat. Oncol Lett 3: 520–524.
  40. 40. Yamamoto M, Kurita A, Asahara T, Takakura A, Katono K, et al. (2008) Metabolism of irinotecan and its active metabolite SN-38 by intestinal microflora in rats. Oncol Rep 20: 727–730.
  41. 41. Kuhn JG (1998) Pharmacology of irinotecan. Oncology (Williston Park) 12: 39–42.
  42. 42. Takahashi H, Honda H (2006) Modified signal-to-noise: a new simple and practical gene filtering approach based on the concept of projective adaptive resonance theory (PART) filtering method. Bioinformatics 22: 1662–1664.
  43. 43. Takahashi H, Kobayashi T, Honda H (2005) Construction of robust prognostic predictors by using projective adaptive resonance theory as a gene filtering method. Bioinformatics 21: 179–186.
  44. 44. Takahashi H, Iwakawa H, Nakao S, Ojio T, Morishita R, et al. (2008) Knowledge-based fuzzy adaptive resonance theory and its application to the analysis of gene expression in plants. J Biosci Bioeng 106: 587–593.
  45. 45. Takahashi H, Honda H (2006) Prediction of peptide binding to major histocompatibility complex class II molecules through use of boosted fuzzy classifier with SWEEP operator method. J Biosci Bioeng 101: 137–141.
  46. 46. Kawamura T, Takahashi H, Honda H (2008) Proposal of new gene filtering method, BagPART, for gene expression analysis with small sample. J Biosci Bioeng 105: 81–84.
  47. 47. Takahashi H, Takahashi A, Naito S, Onouchi H (2012) BAIUCAS: a novel BLAST-based algorithm for the identification of upstream open reading frames with conserved amino acid sequences and its application to the Arabidopsis thaliana genome. Bioinformatics 28: 2231–2241.
  48. 48. Chiba Y, Mineta K, Hirai MY, Suzuki Y, Kanaya S, et al. (2013) Changes in mRNA stability associated with cold stress in Arabidopsis cells. Plant Cell Physiol 54: 180–194.
  49. 49. Iwasaki M, Takahashi H, Iwakawa H, Nakagawa A, Ishikawa T, et al. (2013) Dual regulation of ETTIN (ARF3) gene expression by AS1-AS2, which maintains the DNA methylation level, is involved in stabilization of leaf adaxial-abaxial partitioning in Arabidopsis. Development 140: 1958–1969.
  50. 50. Kojima S, Iwasaki M, Takahashi H, Imai T, Matsumura Y, et al. (2011) ASYMMETRIC LEAVES2 and Elongator, a histone acetyltransferase complex, mediate the establishment of polarity in leaves of Arabidopsis thaliana. Plant Cell Physiol 52: 1259–1273.
  51. 51. Kotooka N, Komatsu A, Takahashi H, Nonaka M, Kawaguchi C, et al. (2013) Predictive value of high-molecular weight adiponectin in subjects with a higher risk of the development of metabolic syndrome: From a population based 5-year follow-up data. Int J Cardiol 167: 1068–1070.
  52. 52. Matsuo N, Mase H, Makino M, Takahashi H, Banno H (2009) Identification of ENHANCER OF SHOOT REGENERATION 1-upregulated genes during in vitro shoot regeneration. Plant Biotechnol 26: 385–393.
  53. 53. Nakagawa A, Takahashi H, Kojima S, Sato N, Ohga K, et al. (2012) Berberine enhances defects in the establishment of leaf polarity in asymmetric leaves1 and asymmetric leaves2 of Arabidopsis thaliana. Plant Mol Biol 79: 569–581.
  54. 54. Nakayama R, Nemoto T, Takahashi H, Ohta T, Kawai A, et al. (2007) Gene expression analysis of soft tissue sarcomas: characterization and reclassification of malignant fibrous histiocytoma. Mod Pathol 20: 749–759.
  55. 55. Sano M, Aoyagi K, Takahashi H, Kawamura T, Mabuchi T, et al. (2010) Forkhead box A1 transcriptional pathway in KRT7-expressing esophageal squamous cell carcinomas with extensive lymph node metastasis. Int J Oncol 36: 321–330.
  56. 56. Yajima I, Kumasaka MY, Naito Y, Yoshikawa T, Takahashi H, et al. (2012) Reduced GNG2 expression levels in mouse malignant melanomas and human melanoma cell lines. Am J Cancer Res 2: 322–329.
  57. 57. Yoshimura K, Mori T, Yokoyama K, Koike Y, Tanabe N, et al. (2011) Identification of alternative splicing events regulated by an Arabidopsis serine/arginine-like protein, atSR45a, in response to high-light stress using a tiling array. Plant Cell Physiol 52: 1786–1805.
  58. 58. Takahashi H, Iwakawa H, Ishibashi N, Kojima S, Matsumura Y, et al. (2013) Meta-analyses of microarrays of arabidopsis asymmetric leaves1 (as1), as2 and their modifying mutants reveal a critical role for the ETT pathway in stabilization of adaxial-abaxial patterning and cell division during leaf development. Plant Cell Physiol 54: 418–431.
  59. 59. Takahashi H, Murase Y, Kobayashi T, Honda H (2007) New cancer diagnosis modeling using boosting and projective adaptive resonance theory with improved reliable index. Biochem Eng J 33: 100–109.
  60. 60. Takahashi H, Nemoto T, Yoshida T, Honda H, Hasegawa T (2006) Cancer diagnosis marker extraction for soft tissue sarcomas based on gene expression profiling data by using projective adaptive resonance theory (PART) filtering method. BMC Bioinformatics 7: 399.
  61. 61. Takahashi H, Aoyagi K, Nakanishi Y, Sasaki H, Yoshida T, et al. (2006) Classification of intramural metastases and lymph node metastases of esophageal cancer from gene expression based on boosting and projective adaptive resonance theory. J Biosci Bioeng 102: 46–52.
  62. 62. Takahashi H, Honda H (2006) Lymphoma prognostication from expression profiling using a combination method of boosting and projective adaptive resonance theory. J Chem Eng Jpn 39: 767–771.
  63. 63. Takahashi H, Honda H (2005) A new reliable cancer diagnosis method using boosted fuzzy classifier with a SWEEP operator method. J Chem Eng Jpn 38: 763–773.
  64. 64. Takahashi H, Masuda K, Ando T, Kobayashi T, Honda H (2004) Prognostic predictor with multiple fuzzy neural models using expression profiles from DNA microarray for metastases of breast cancer. J Biosci Bioeng 98: 193–199.
  65. 65. Takahashi H, Tomida S, Kobayashi T, Honda H (2003) Inference of common genetic network using fuzzy adaptive resonance theory associated matrix method. J Biosci Bioeng 96: 154–160.
  66. 66. Takahashi H, Nakayama R, Hayashi S, Nemoto T, Murase Y, et al. (2013) Macrophage migration inhibitory factor and stearoyl-CoA desaturase 1: potential prognostic markers for soft tissue sarcomas based on bioinformatics analyses. PLoS One 8: e78250.
  67. 67. Fisher RA (1922) On the interpretation of χ2 from contingency tables, and the calculation of P. J Roy Statistical Society 85: 87–94.
  68. 68. Pitman EJG (1938) Significance tests which may be applied to samples from any population. Part III. The analysis of variance test. Biometrika 29: 322–335.
  69. 69. Benjamini Y, Hochberg Y (1995) Controlling the false discovery rate: a practical and powerful approach to multiple testing. J R Statist Soc serB 57: 298–300.
  70. 70. Akaike H (1974) A new look at the statistical model identification. IEEE T Automat Contr 19: 716–723.
  71. 71. Sakamoto Y, Ishiguro M, Kitagawa G (1986) Akaike Information Criterion Statistics. Dordrecht: Reidel Publishing Company.
  72. 72. Skol AD, Scott LJ, Abecasis GR, Boehnke M (2006) Joint analysis is more efficient than replication-based analysis for two-stage genome-wide association studies. Nat Genet 38: 209–213.
  73. 73. Delmas P, Brown DA (2005) Pathways modulating neural KCNQ/M (Kv7) potassium channels. Nat Rev Neurosci 6: 850–862.
  74. 74. Miceli F, Soldovieri MV, Martire M, Taglialatela M (2008) Molecular pharmacology and therapeutic potential of neuronal Kv7-modulating drugs. Curr Opin Pharmacol 8: 65–74.
  75. 75. Brown DA, Passmore GM (2009) Neural KCNQ (Kv7) channels. Br J Pharmacol 156: 1185–1195.
  76. 76. Jepps TA, Greenwood IA, Moffatt JD, Sanders KM, Ohya S (2009) Molecular and functional characterization of Kv7 K+ channel in murine gastrointestinal smooth muscles. Am J Physiol Gastrointest Liver Physiol 297: G107–115.
  77. 77. Schwake M, Jentsch TJ, Friedrich T (2003) A carboxy-terminal domain determines the subunit specificity of KCNQ K+ channel assembly. EMBO Rep 4: 76–81.
  78. 78. Cavaliere S, Malik BR, Hodge JJ (2013) KCNQ channels regulate age-related memory impairment. PLoS One 8: e62445.
  79. 79. Perez C, Vega R, Soto E (2010) Phospholipase C-mediated inhibition of the M-potassium current by muscarinic-receptor activation in the vestibular primary-afferent neurons of the rat. Neurosci Lett 468: 238–242.
  80. 80. Blandizzi C, De Paolis B, Colucci R, Lazzeri G, Baschiera F, et al. (2001) Characterization of a novel mechanism accounting for the adverse cholinergic effects of the anticancer drug irinotecan. Br J Pharmacol 132: 73–84.
  81. 81. Yasuda K, Miyake K, Horikawa Y, Hara K, Osawa H, et al. (2008) Variants in KCNQ1 are associated with susceptibility to type 2 diabetes mellitus. Nat Genet 40: 1092–1097.
  82. 82. Ichihara S, Toyooka S, Fujiwara Y, Hotta K, Shigematsu H, et al. (2007) The impact of epidermal growth factor receptor gene status on gefitinib-treated Japanese patients with non-small-cell lung cancer. Int J Cancer 120: 1239–1247.
  83. 83. Seo S, Takayama K, Uno K, Ohi K, Hashimoto R, et al. (2013) Functional Analysis of Deep Intronic SNP rs13438494 in Intron 24 of PCLO Gene. PLoS One 8: e76960.
  84. 84. Sauna ZE, Kimchi-Sarfaty C, Ambudkar SV, Gottesman MM (2007) Silent polymorphisms speak: how they affect pharmacogenomics and the treatment of cancer. Cancer Res 67: 9609–9612.
  85. 85. Capon F, Allen MH, Ameen M, Burden AD, Tillman D, et al. (2004) A synonymous SNP of the corneodesmosin gene leads to increased mRNA stability and demonstrates association with psoriasis across diverse ethnic groups. Hum Mol Genet 13: 2361–2368.
  86. 86. Nackley AG, Shabalina SA, Tchivileva IE, Satterfield K, Korchynskyi O, et al. (2006) Human catechol-O-methyltransferase haplotypes modulate protein expression by altering mRNA secondary structure. Science 314: 1930–1933.
  87. 87. Nielsen KB, Sorensen S, Cartegni L, Corydon TJ, Doktor TK, et al. (2007) Seemingly neutral polymorphic variants may confer immunity to splicing-inactivating mutations: a synonymous SNP in exon 5 of MCAD protects from deleterious mutations in a flanking exonic splicing enhancer. Am J Hum Genet 80: 416–432.
  88. 88. Spasovski V, Tosic N, Nikcevic G, Stojiljkovic M, Zukic B, et al. (2013) The influence of novel transcriptional regulatory element in intron 14 on the expression of Janus kinase 2 gene in myeloproliferative neoplasms. J Appl Genet 54: 21–26.
  89. 89. Xue G, Aida Y, Onodera T, Sakudo A (2012) The 5′ flanking region and intron1 of the bovine prion protein gene (PRNP) are responsible for negative feedback regulation of the prion protein. PLoS One 7: e32870.
  90. 90. Yoshida T, Ono H, Kuchiba A, Saeki N, Sakamoto H (2010) Genome-wide germline analyses on cancer susceptibility and GeMDBJ database: Gastric cancer as an example. Cancer Sci 101: 1582–1589.