[Show abstract][Hide abstract] ABSTRACT: Background:
Copy number aberrations frequently occur during the development of many cancers. Such events affect dosage of involved genes and may cause further genomic instability and progression of cancer. In this survey, canine SNP microarrays were used to study 117 canine mammary tumours from 69 dogs.
We found a high occurrence of copy number aberrations in canine mammary tumours, losses being more frequent than gains. Increased frequency of aberrations and loss of heterozygosity were positively correlated with increased malignancy in terms of histopathological diagnosis. One of the most highly recurrently amplified regions harbored the MYC gene. PTEN was located to a frequently lost region and also homozygously deleted in five tumours. Thus, deregulation of these genes due to copy number aberrations appears to be an important event in canine mammary tumour development. Other potential contributors to canine mammary tumour pathogenesis are COL9A3, INPP5A, CYP2E1 and RB1. The present study also shows that a more detailed analysis of chromosomal aberrations associated with histopathological parameters may aid in identifying specific genes associated with canine mammary tumour progression.
The high frequency of copy number aberrations is a prominent feature of canine mammary tumours as seen in other canine and human cancers. Our findings share several features with corresponding studies in human breast tumours and strengthen the dog as a suitable model organism for this disease.
PLoS ONE 05/2015; 10(5):e0126371. DOI:10.1371/journal.pone.0126371 · 3.23 Impact Factor
[Show abstract][Hide abstract] ABSTRACT: Breast cancer is a heterogeneous disease for which alterations in DNA methylation patterns have been shown to be of biological and clinical importance. Here we report on the integrated analysis of molecular alterations including the methylation status of 27 gene promoters analyzed by highly quantitative pyrosequencing, and the association to gene expression, germline genotype and clinical parameters including survival. Breast cancer specific deregulation of DNA methylation (both hyper- and hypomethylation) was found in twenty genes including ACVR1, OGG1, IL8 and TFF1. The methylation level in the promoter regions was significantly negatively correlated to gene expression for twelve genes (such as MST1R, ST6GAL1 and TFF1) indicating that a gain of aberrant methylation (hypermethylation) inhibits gene expression. Multiple associations between molecular and clinical parameters were identified, and multivariate statistical analysis demonstrated that methylation was more strongly associated to clinical parameters than gene expression for the investigated genes. The methylation level of BCAP31 and OGG1 showed significant association to survival, and these associations were validated in a larger patient cohort (The Cancer Genome Atlas). Our study provides evidence for the promise of DNA methylation alterations for clinical applications.
International Journal of Cancer 06/2014; 134(11). DOI:10.1002/ijc.28606 · 5.09 Impact Factor
[Show abstract][Hide abstract] ABSTRACT: Aberrant DNA methylation of regulatory genes has frequently been found in human breast cancers and correlated to clinical outcome. In the present study we investigate stage specific changes in the DNA methylation patterns in order to identify valuable markers to understand how these changes affect breast cancer progression.
Quantitative DNA methylation analyses of 12 candidate genes ABCB1, BRCCA1, CDKN2A, ESR1, GSTP1, IGF2, MGMT, HMLH1, PPP2R2B, PTEN, RASSF1A and FOXC1 was performed by pyrosequencing a series of 238 breast cancer tissue samples from DCIS to invasive tumors stage I to IV.
Significant differences in methylation levels between the DCIS and invasive stage II tumors were observed for six genes RASSF1A, CDKN2A, MGMT, ABCB1, GSTP1 and FOXC1. RASSF1A, ABCB1 and GSTP1 showed significantly higher methylation levels in late stage compared to the early stage breast carcinoma. Z-score analysis revealed significantly lower methylation levels in DCIS and stage I tumors compared with stage II, III and IV tumors. Methylation levels of PTEN, PPP2R2B, FOXC1, ABCB1 and BRCA1 were lower in tumors harboring TP53 mutations then in tumors with wild type TP53. Z-score analysis showed that TP53 mutated tumors had significantly lower overall methylation levels compared to tumors with wild type TP53. Methylation levels of RASSF1A, PPP2R2B, GSTP1 and FOXC1 were higher in ER positive vs. ER negative tumors and methylation levels of PTEN and CDKN2A were higher in HER2 positive vs. HER2 negative tumors. Z-score analysis also showed that HER2 positive tumors had significantly higher z-scores of methylation compared to the HER2 negative tumors. Univariate survival analysis identifies methylation status of PPP2R2B as significant predictor of overall survival and breast cancer specific survival.
In the present study we report that the level of aberrant DNA methylation is higher in late stage compared with early stage of invasive breast cancers and DCIS for genes mentioned above.
BMC Cancer 10/2013; 13(1):456. DOI:10.1186/1471-2407-13-456 · 3.36 Impact Factor
[Show abstract][Hide abstract] ABSTRACT: Breast cancers today are of predominantly T1 (0.1 ≥ 2.0 cm) or T2 (>2 ≤ 5 cm) categories due to early diagnosis. Molecular profiling using microarrays has led to the notion of breast cancer as a heterogeneous disease both clinically and molecularly. Given the prognostic power and clinical use of tumor size, the purpose of this study was to search for molecular signatures characterizing clinical T1 and T2. In total 46 samples were included in the discovery dataset. After adjusting for hormone receptor status, lymph node status, grade, and tumor subclass 441 genes were differently expressed between T1 and T2 tumors. Focal adhesion and extracellular matrix receptor interaction were upregulated in the smaller tumors while p38MAPK signaling and immune-related pathways were more dominant in the larger tumors.
The T-size signature was then tested on a validation set of 947 breast tumor samples. Using the T-size expression signatures instead of tumor size leads to a significant difference in risk for distant metastases (P < 0.001). If further confirmed, this molecular signature can be used to select patients with tumor category T1 who may need more aggressive treatment and patients with tumor category T2 who may have less benefit from it.
[Show abstract][Hide abstract] ABSTRACT: Background
Recent studies have reported associations between a variant allele in a let-7 microRNA complementary site (LCS6) within the 3′untranslated region (3′UTR) of KRAS (rs61764370) and clinical outcome in metastatic colorectal cancer (mCRC) patients receiving cetuximab. The variant allele has also been associated with increased cancer risk. We aimed to reveal the incidence of the variant allele in a colorectal cancer screening population and to investigate the clinical relevance of the variant allele in mCRC patients treated with 1st line Nordic FLOX (bolus 5-fluorouracil/folinic acid and oxaliplatin) +/− cetuximab.
The feasibility of the variant allele as a risk factor for CRC was investigated by comparing the LCS6 gene frequencies in 197 CRC patients, 1060 individuals with colorectal polyps, and 358 healthy controls. The relationship between clinical outcome and LCS6 genotype was analyzed in 180 mCRC patients receiving Nordic FLOX and 355 patients receiving Nordic FLOX + cetuximab in the NORDIC-VII trial (NCT00145314).
LCS6 frequencies did not vary between CRC patients (23%), individuals with polyps (20%), and healthy controls (20%) (P = 0.50). No statistically significant differences were demonstrated in the NORDIC-VII cohort even if numerically increased progression-free survival (PFS) and overall survival (OS) were found in patients with the LCS6 variant allele (8.5 (95% CI: 7.3-9.7 months) versus 7.8 months (95% CI: 7.4-8.3 months), P = 0.16 and 23.5 (95% CI: 21.6-25.4 months) versus 19.5 months (95% CI: 17.8-21.2 months), P = 0.31, respectively). Addition of cetuximab seemed to improve response rate more in variant carriers than in wild-type carriers (from 35% to 57% versus 44% to 47%), however the difference was not statistically significant (interaction P = 0.16).
The LCS6 variant allele does not seem to be a risk factor for development of colorectal polyps or CRC. No statistically significant effect of the LCS6 variant allele on response rate, PFS or OS was found in mCRC patients treated with 1st line Nordic FLOX +/− cetuximab.
BMC Cancer 11/2012; 12(1):534. DOI:10.1186/1471-2407-12-534 · 3.36 Impact Factor
[Show abstract][Hide abstract] ABSTRACT: The majority of human breast cancers exhibit luminal epithelial differentiation. However, most aggressive behavior, including invasion and purported cancer stem cell activity, are considered characteristics of basal-like cells. We asked the following questions: Must luminal-like breast cancer cells become basal-like to initiate tumors or to invade? Could luminally differentiated cells within a basally initiated hierarchy also be tumorigenic? To answer these questions, we used rare and mutually exclusive lineage markers to isolate subsets of luminal-like and basal-like cells from human breast tumors. We enriched for populations with or without prominent basal-like traits from individual tumors or single cell cloning from cell lines and recovered cells with a luminal-like phenotype. Tumor cells with basal-like traits mimicked phenotypic and functional behavior associated with stem cells assessed by gene expression, mammosphere formation and lineage markers. Luminal-like cells without basal-like traits, surprisingly, were fully capable of initiating invasive tumors in NOD SCID gamma (NSG) mice. In fact, these phenotypically pure luminal-like cells generated larger and more invasive tumors than their basal-like counterparts. The tumorigenicity and invasive potential of the luminal-like cancer cells relied strongly on the expression of the gene GCNT1, which encodes a key glycosyltransferase controlling O-glycan branching. These findings demonstrate that basal-like cells, as defined currently, are not a requirement for breast tumor aggressiveness, and that within a single tumor there are multiple "stem-like" cells with tumorigenic potential casting some doubt on the hypothesis of hierarchical or differentiative loss of tumorigenicity.
Proceedings of the National Academy of Sciences 03/2012; 109(16):6124-9. DOI:10.1073/pnas.1203203109 · 9.67 Impact Factor
[Show abstract][Hide abstract] ABSTRACT: In a cancer cell the number of copies of a locus may vary due to amplification and deletion and these variations are denoted as copy number alterations (CNAs). We focus on the disparity of CNAs in tumour samples, which were compared to those in blood in order to identify the directional loss of heterozygosity.
We propose a numerical algorithm and apply it to data from the Illumina 109K-SNP array on 112 samples from breast cancer patients. B-allele frequency (BAF) and log R ratio (LRR) of Illumina were used to estimate Euclidian distances. For each locus, we compared genotypes in blood and tumour for subset of samples being heterozygous in blood. We identified loci showing preferential disparity from heterozygous toward either the A/B-allele homozygous (allelic disparity). The chi-squared and Cochran-Armitage trend tests were used to examine whether there is an association between high levels of disparity in single nucleotide polymorphisms (SNPs) and molecular, clinical and tumour-related parameters. To identify pathways and network functions over-represented within the resulting gene sets, we used Ingenuity Pathway Analysis (IPA).
To identify loci with a high level of disparity, we selected SNPs 1) with a substantial degree of disparity and 2) with substantial frequency (at least 50% of the samples heterozygous for the respective locus). We report the overall difference in disparity in high-grade tumours compared to low-grade tumours (p-value < 0.001) and significant associations between disparity in multiple single loci and clinical parameters. The most significantly associated network functions within the genes represented in the loci of disparity were identified, including lipid metabolism, small-molecule biochemistry, and nervous system development and function. No evidence for over-representation of directional disparity in a list of stem cell genes was obtained, however genes appeared to be more often altered by deletion than by amplification.
Our data suggest that directional loss and amplification exist in breast cancer. These are highly associated with grade, which may indicate that they are enforced with increasing number of cell divisions. Whether there is selective pressure for some loci to be preferentially amplified or deleted remains to be confirmed.
BMC Medical Genomics 12/2011; 4:85. DOI:10.1186/1755-8794-4-85 · 2.87 Impact Factor
[Show abstract][Hide abstract] ABSTRACT: Elucidating the exact relationship between gene copy number and expression would enable identification of regulatory mechanisms of abnormal gene expression and biological pathways of regulation. Most current approaches either depend on linear correlation or on nonparametric tests of association that are insensitive to the exact shape of the relationship. Based on knowledge of enzyme kinetics and gene regulation, we would expect the functional shape of the relationship to be gene dependent and to be related to the gene regulatory mechanisms involved. Here, we propose a statistical approach to investigate and distinguish between linear and nonlinear dependences between DNA copy number alteration and mRNA expression.
We applied the proposed method to DNA copy numbers derived from Illumina 109 K SNP-CGH arrays (using the log R values) and expression data from Agilent 44 K mRNA arrays, focusing on commonly aberrated genomic loci in a collection of 102 breast tumors. Regression analysis was used to identify the type of relationship (linear or nonlinear), and subsequent pathway analysis revealed that genes displaying a linear relationship were overall associated with substantially different biological processes than genes displaying a nonlinear relationship. In the group of genes with a linear relationship, we found significant association to canonical pathways, including purine and pyrimidine metabolism (for both deletions and amplifications) as well as estrogen metabolism (linear amplification) and BRCA-related response to damage (linear deletion). In the group of genes displaying a nonlinear relationship, the top canonical pathways were specific pathways like PTEN and PI13K/AKT (nonlinear amplification) and Wnt(B) and IL-2 signalling (nonlinear deletion). Both amplifications and deletions pointed to the same affected pathways and identified cancer as the top significant disease and cell cycle, cell signaling and cellular development as significant networks.
This paper presents a novel approach to assessing the validity of the dependence of expression data on copy number data, and this approach may help in identifying the drivers of carcinogenesis.
[Show abstract][Hide abstract] ABSTRACT: Breast cancer is a heterogeneous disease that can be divided in subtypes based on histology, gene expression profiles as well as differences in genomic aberrations. Distinct global DNA methylation profiles have been reported in normal breast epithelial cells as well as in breast tumors. However, the influence of the tumor methylome on the previously described subgroups of breast cancer is not fully understood. Here we report the DNA methylation profiles of 80 breast tumors using a panel of 807 cancer related genes interrogating 1505 CpG sites. We identified three major clusters based on the methylation profiles; one consisting of mainly tumors of myoepithelial origin and two other clusters with tumors of predominantly luminal epithelial origin. The clusters were different with respect to estrogen receptor status, TP53 status, ErbB2 status and grade. The most significantly differentially methylated genes including HDAC1, TFF1, OGG1, BMP3, FZD9 and HOXA11 were confirmed by pyrosequencing. Gene Ontology analysis revealed enrichment for genes involved in developmental processes including homeobox domain genes (HOXA9, HOXA11, PAX6, MYBL2, ISL1 and IPF1) and (ETS1, HDAC1, CREBBP, GAS7, SPI1 and TBX1). Extensive correlation to mRNA expression was observed. Pathway analyses identified a significant association with canonical (curated) pathways such as hepatic fibrosis including genes like EGF, NGFR and TNF, dendritic cell maturation and the NF-κB signaling pathway. Our results show that breast tumor expression subtypes harbor major epigenetic differences and tumors with similar gene expression profiles might belong to epigenetically different subtypes. Some of the transcription factors identified, with key roles in differentiation and development might play a role in inducing and maintaining the different phenotypes.
[Show abstract][Hide abstract] ABSTRACT: Mammographic density (MD), as assessed from film screen mammograms, is determined by the relative content of adipose, connective and epithelial tissue in the female breast. In epidemiological studies, a high percentage of MD confers a four to six fold risk elevation of developing breast cancer, even after adjustment for other known breast cancer risk factors. However, the biologic correlates of density are little known.
Gene expression analysis using whole genome arrays was performed on breast biopsies from 143 women; 79 women with no malignancy (healthy women) and 64 newly diagnosed breast cancer patients, both included from mammographic centres. Percent MD was determined using a previously validated, computerized method on scanned mammograms. Significance analysis of microarrays (SAM) was performed to identify genes influencing MD and a linear regression model was used to assess the independent contribution from different variables to MD.
SAM-analysis identified 24 genes differentially expressed between samples from breasts with high and low MD. These genes included three uridine 5'-diphospho-glucuronosyltransferase (UGT) genes and the oestrogen receptor gene (ESR1). These genes were down-regulated in samples with high MD compared to those with low MD. The UGT gene products, which are known to inactivate oestrogen metabolites, were also down-regulated in tumour samples compared to samples from healthy individuals. Several single nucleotide polymorphisms (SNPs) in the UGT genes associated with the expression of UGT and other genes in their vicinity were identified.
Three UGT enzymes were lower expressed both in breast tissue biopsies from healthy women with high MD and in biopsies from newly diagnosed breast cancers. The association was strongest amongst young women and women using hormonal therapy. UGT2B10 predicts MD independently of age, hormone therapy and parity. Our results indicate that down-regulation of UGT genes in women exposed to female sex hormones is associated with high MD and might increase the risk of breast cancer.
Breast cancer research: BCR 08/2010; 12(4):R65. DOI:10.1186/bcr2632 · 5.49 Impact Factor
[Show abstract][Hide abstract] ABSTRACT: Breast cancer is the most frequent cancer in women and consists of a heterogeneous collection of diseases with distinct histopathological, genetic and epigenetic characteristics. In this study, we aimed to identify DNA methylation based biomarkers to distinguish patients with locally advanced breast cancer who may benefit from neoadjuvant doxorubicin treatment.
We investigated quantitatively the methylation patterns in the promoter regions of 14 genes (ABCB1, ATM, BRCA1, CDH3, CDKN2A, CXCR4, ESR1, FBXW7, FOXC1, GSTP1, IGF2, HMLH1, PPP2R2B, and PTEN) in 75 well-described pre-treatment samples from locally advanced breast cancer and correlated the results to the available clinical and molecular parameters. Six normal breast tissues were used as controls and 163 unselected breast cancer cases were used to validate associations with histopathological and clinical parameters.Aberrant methylation was detected in 9 out of the 14 genes including the discovery of methylation at the FOXC1 promoter. Absence of methylation at the ABCB1 promoter correlated with progressive disease during doxorubicin treatment. Most importantly, the DNA methylation status at the promoters of GSTP1, FOXC1 and ABCB1 correlated with survival, whereby the combination of methylated genes improved the subdivision with respect to the survival of the patients. In multivariate analysis GSTP1 and FOXC1 methylation status proved to be independent prognostic markers associated with survival.
Quantitative DNA methylation profiling is a powerful tool to identify molecular changes associated with specific phenotypes. Methylation at the ABCB1 or GSTP1 promoter improved overall survival probably due to prolonged availability and activity of the drug in the cell while FOXC1 methylation might be a protective factor against tumour invasiveness. FOXC1 proved to be general prognostic factor, while ABCB1 and GSTP1 might be predictive factors for the response to and efficacy of doxorubicin treatment. Pharmacoepigenetic effects such as the reported associations in this study provide molecular explanations for differential responses to chemotherapy and it might prove valuable to take the methylation status of selected genes into account for patient management and treatment decisions.
Molecular Cancer 03/2010; 9(1):68. DOI:10.1186/1476-4598-9-68 · 4.26 Impact Factor
[Show abstract][Hide abstract] ABSTRACT: Increased mammographic density is one of the strongest independent risk factors for breast cancer. It is believed that one third of breast cancers are derived from breasts with more than 50% density. Mammographic density is affected by age, BMI, parity, and genetic predisposition. It is also greatly influenced by hormonal and growth factor changes in a woman's life cycle, spanning from puberty through adult to menopause. Genetic variations in genes coding for hormones and growth factors involved in development of the breast are therefore of great interest. The associations between genetic polymorphisms in genes from the IGF pathway on mammographic density and circulating levels of IGF1, its binding protein IGFBP3, and their ratio in postmenopausal women are reported here.
Samples from 964 postmenopausal Norwegian women aged 55-71 years were collected as a part of the Tromsø Mammography and Breast Cancer Study. All samples were genotyped for 25 SNPs in IGF1, IGF2, IGF1R, IGF2R, IGFALS and IGFBP3 using Taqman (ABI). The main statistical analyses were conducted with the PROC HAPLOTYPE procedure within SAS/GENETICS (SAS 9.1.3).
The haplotype analysis revealed six haploblocks within the studied genes. Of those, four had significant associations with circulating levels of IGF1 or IGFBP3 and/or mammographic density. One haplotype variant in the IGF1 gene was found to be associated with mammographic density. Within the IGF2 gene one haplotype variant was associated with levels of both IGF1 and IGFBP3. Two haplotype variants in the IGF2R were associated with the level of IGF1. Both variants of the IGFBP3 haplotype were associated with the IGFBP3 level and indicate regulation in cis.
Polymorphisms within the IGF1 gene and related genes were associated with plasma levels of IGF1, IGFBP3 and mammographic density in this study of postmenopausal women.
BMC Medical Genomics 03/2010; 3(1):9. DOI:10.1186/1755-8794-3-9 · 2.87 Impact Factor
[Show abstract][Hide abstract] ABSTRACT: The dose of docetaxel is currently calculated based on body surface area and does not reflect the pharmacokinetic, metabolic potential or genetic background of the patients. The influence of genetic variation on the clearance of docetaxel was analysed in a two-stage analysis. In step one, 583 single-nucleotide polymorphisms (SNPs) in 203 genes were genotyped on samples from 24 patients with locally advanced non-small cell lung cancer. We found that many of the genes harbour several SNPs associated with clearance of docetaxel. Most notably these were four SNPs in EGF, three SNPs in PRDX4 and XPC, and two SNPs in GSTA4, TGFBR2, TNFAIP2, BCL2, DPYD and EGFR. The multiple SNPs per gene suggested the existence of common haplotypes associated with clearance. These were confirmed with detailed haplotype analysis. On the basis of analysis of variance (ANOVA), quantitative mutual information score (QMIS) and Kruskal-Wallis (KW) analysis SNPs significantly associated with clearance of docetaxel were confirmed for GSTA4, PRDX4, TGFBR2 and XPC and additional putative markers were found in CYP2C8, EPHX1, IGF2, IL1R2, MAPK7, NDUFB4, TGFBR3, TPMT (2 SNPs), (P<0.05 or borderline significant for all three methods, 14 SNPs in total). In step two, these 14 SNPs were genotyped in additional 9 samples and the results combined with the genotyping results from the first step. For 7 of the 14 SNPs, the results are still significant/borderline significant by all three methods: ANOVA, QMIS and KW analysis strengthening our hypothesis that they are associated with the clearance of docetaxel.
The Pharmacogenomics Journal 02/2010; 10(6):513-23. DOI:10.1038/tpj.2010.6 · 4.23 Impact Factor
[Show abstract][Hide abstract] ABSTRACT: Ductal carcinoma in situ (DCIS) is a non-invasive lesion of the breast that is frequently detected by mammography and subsequently removed by surgery. However, it is estimated that about half of the detected lesions would never have progressed into invasive cancer. Identifying DCIS and invasive cancer specific epigenetic lesions and understanding how these epigenetic changes are involved in triggering tumour progression is important for a better understanding of which lesions are at risk of becoming invasive.
Quantitative DNA methylation analysis of ABCB1, CDKN2A/p16INK4a, ESR1, FOXC1, GSTP1, IGF2, MGMT, MLH1, PPP2R2B, PTEN and RASSF1A was performed by pyrosequencing in a series of 27 pure DCIS, 28 small invasive ductal carcinomas (IDCs), 34 IDCs with a DCIS component and 5 normal breast tissue samples. FOXC1, ABCB1, PPP2R2B and PTEN were analyzed in 23 additional normal breast tissue samples. Real-Time PCR expression analysis was performed for FOXC1.
Aberrant DNA methylation was observed in all three diagnosis groups for the following genes: ABCB1, FOXC1, GSTP1, MGMT, MLH1, PPP2R2B, PTEN and RASSF1A. For most of these genes, methylation was already present at the DCIS level with the same frequency as within IDCs. For FOXC1 significant differences in methylation levels were observed between normal breast tissue and invasive tumours (P < 0.001). The average DNA methylation levels were significantly higher in the pure IDCs and IDCs with DCIS compared to pure DCIS (P = 0.007 and P = 0.001, respectively). Real-time PCR analysis of FOXC1 expression from 25 DCIS, 23 IDCs and 28 normal tissue samples showed lower gene expression levels of FOXC1 in both methylated and unmethylated tumours compared to normal tissue (P < 0.001). DNA methylation levels of FOXC1, GSTP1, ABCB1 and RASSF1A were higher in oestrogen receptor (ER) positive vs. ER negative tumours; whereas methylation levels of FOXC1, ABCB1, PPP2R2B and PTEN were lower in tumours with a TP53 mutation.
Quantitative methylation analysis identified ABCB1, FOXC1, PPP2R2B and PTEN as novel genes to be methylated in DCIS. In particular, FOXC1 showed a significant increase in the methylation frequency in invasive tumours. Low FOXC1 gene expression in both methylated and unmethylated DCIS and IDCs indicates that the loss of its expression is an early event during breast cancer progression.
Breast cancer research: BCR 01/2010; 12(1):R3. DOI:10.1186/bcr2466 · 5.49 Impact Factor
[Show abstract][Hide abstract] ABSTRACT: In frequency-domain blind source separation (BSS) for speech with independent component analysis (ICA), a practical parametric Pearson distribution system is used to model the distribution of frequency-domain source signals. ICA adaptation rules have a score function determined by an approximated signal distribution. Approximation based on the data may produce better separation performance than we can obtain with ICA. Previously, conventional hyperbolic tangent (tanh) or generalized Gaussian distribution (GGD) was uniformly applied to the score function for all frequency bins, even though a wideband speech signal has different distributions at different frequencies. To deal with this, we propose modeling the signal distribution at each frequency by adopting a parametric Pearson distribution and employing it to optimize the separation matrix in the ICA learning process. The score function is estimated by the appropriate Pearson distribution parameters for each frequency bin. We devised three methods for Pearson distribution parameter estimation and conducted separation experiments with real speech signals convolved with actual room impulse responses (T<sub>60</sub>=130 ms). Our experimental results show that the proposed frequency-domain Pearson-ICA (FD-Pearson-ICA) adapted well to the characteristics of frequency-domain source signals. By applying the FD-Pearson-ICA performance, the signal-to-interference ratio significantly improved by around 2-3 dB compared with conventional nonlinear functions. Even if the signal-to-interference ratio (SIR) values of FD-Pearson-ICA were poor, the performance based on a disparity measure between the true score function and estimated parametric score function clearly showed the advantage of FD-Pearson-ICA. Furthermore, we confirmed the optimum of the proposed approach for/optimized the proposed approach as regards separation performance. By combining individual distribution parameters directly estimated at low frequency with the ap-
propriate parameters optimized at high frequency, it was possible to both reasonably improve the FD-Pearson-ICA performance without any significant increase in the computational burden by comparison with conventional nonlinear functions.
IEEE Transactions on Audio Speech and Language Processing 06/2009; 17(4-17):639 - 649. DOI:10.1109/TASL.2008.2011527 · 2.48 Impact Factor
[Show abstract][Hide abstract] ABSTRACT: The CpG island spanning the transcription start of the glutathione S-transferase P1 becomes methylated in a variety of human cancers including breast cancer. To study the effect of sequence variation on hypermethylation of the GSTP1 promoter, we analyzed the genetic and epigenetic variability in 90 tumors from patients with locally advanced breast cancer. High-resolution quantitative analysis revealed large variability in the DNA methylation levels. Lack of methylation was more often observed in the basal and normal-like estrogen receptor (ER)-negative tumors, and methylated GSTP1 was associated with better overall survival (P = 0.00063). Studies of the genetic variation identified 14 different haplotypes. The distribution of methylation levels of tumors homozygous for the most frequent haplotype was significantly different from other haplotype combinations (P = 0.011), the difference being more pronounced in ER-positive (P = 0.005) and progesterone receptor-positive (P = 0.008) tumors. Regression modeling identified the ER status and haplotype as the main determinants of DNA methylation variability. We identified a putative c-Myb response element (MRE) that was present in one of two minimal promoter haplotypes. In vitro analysis showed that c-Myb binds to the MRE, but binding was weakened by the two polymorphisms. Transient cotransfections in luminal-type and basal-like breast cancer cell lines confirmed cell-specific differential binding of c-Myb to the polymorphic sites, leading to a change in the expression from the GSTP1 promoter in vivo. GSTP1 expression was moderately but significantly (P = 0.01) reduced after siRNA-mediated knockdown of c-Myb. Our results indicate that haplotype structure of a promoter is important for the extent of DNA methylation.
Cancer Research 07/2008; 68(14):5562-71. DOI:10.1158/0008-5472.CAN-07-5828 · 9.33 Impact Factor