Shunji Uchimura

Oligonucleotide microarray for prediction of early intrahepatic recurrence of hepatocellular carcinoma after curative resection

BACKGROUND Hepatocellular carcinoma has a poor prognosis because of the high intrahepatic recurrence rate. There are technological limitations to traditional methods such as TNM staging for accurate prediction of recurrence, suggesting that new techniques are needed. METHODS We investigated mRNA expression profiles in tissue specimens from a training set, comprising 33 patients with hepatocellular carcinoma, with high-density oligonucleotide microarrays representing about 6000 genes. We used this training set in a supervised learning manner to construct a predictive system, consisting of 12 genes, with the Fisher linear classifier. We then compared the predictive performance of our system with that of a predictive system with a support vector machine (SVM-based system) on a blinded set of samples from 27 newly enrolled patients. FINDINGS Early intrahepatic recurrence within 1 year after curative surgery occurred in 12 (36%) and eight (30%) patients in the training and blinded sets, respectively. Our system correctly predicted early intrahepatic recurrence or non-recurrence in 25 (93%) of 27 samples in the blinded set and had a positive predictive value of 88% and a negative predictive value of 95%. By contrast, the SVM-based system predicted early intrahepatic recurrence or non-recurrence correctly in only 16 (60%) individuals in the blinded set, and the result yielded a positive predictive value of only 38% and a negative predictive value of 79%. INTERPRETATION Our system predicted early intrahepatic recurrence or non-recurrence for patients with hepatocellular carcinoma much more accurately than the SVM-based system, suggesting that our system could serve as a new method for characterising the metastatic potential of hepatocellular carcinoma.

A gabor filter-based method for recognizing handwritten numerals

Abstract We study a Gabor-filter-based method for handwritten numeral character recognition. The Gabor filter is based on a multi-channel filtering theory for processing visual information in the early stages of the human visual systems. The performance of the Gabor-filter-based method is demonstrated on the ETL-1 database. Experimental results show that the artificial neural-network classifier achieved the error rate of 2.34% for a test set of 7000 characters. Therefore, the Gabor-filter-based method should be considered in recognition of handwritten numeric characters.

A bootstrap technique for nearest neighbor classifier design

A bootstrap technique for nearest neighbor classifier design is proposed. Our primary interest in designing a classifier is in small training sample size situations. Conventional bootstrapping techniques sample the training samples with replacement. On the other hand, our technique generates bootstrap samples by locally combining original training samples. The nearest neighbor classifier is designed on the bootstrap samples and is tested on the test samples independent of training samples. The performance of the proposed classifier is demonstrated on three artificial data sets and one real data set. Experimental results show that the nearest neighbor classifier designed on the bootstrap samples outperforms the conventional k-NN classifiers as well as the edited 1-NN classifiers, particularly in high dimensions.

Differential gene expression in distinct virologic types of hepatocellular carcinoma: association with liver cirrhosis

Using oligonucleotide microarray data of 45 hepatocellular carcinoma (HCC) samples, we evaluated gene expression in hepatitis B virus-positive and hepatitis C virus-positive HCCs (HBV- and HCV-HCCs) for an association with liver cirrhosis (LC). In all, 89 genes were expressed differentially between HBV-HCCs associated with LC and those not associated with LC. Among them, tumors from LC patients showed significantly lower expression levels of 72 genes and significantly higher levels of 17 genes than the levels found in tumors from non-LC patients. The former included genes responsible for signal transduction, transcription, metabolism, and cell growth. The latter included a tumor suppressor gene and a cell-growth-related gene. Only eight genes were expressed differentially between HCV-HCCs associated with and without LC. Our findings provide as a framework for clarifying the role of LC in HBV- and HCV-related hepatocarcinogenesis.

Fibroblast growth factor 2 facilitates the differentiation of transplanted bone marrow cells into hepatocytes

We have developed an in vivo mouse model, the green fluorescent protein (GFP)/carbon tetrachloride (CCl4) model, and have previously reported that transplanted GFP-positive bone marrow cells (BMCs) differentiate into hepatocytes via hepatoblast intermediates. Here, we have investigated the growth factors that are closely related to the differentiation of transplanted BMCs into hepatocytes, and the way that a specific growth factor affects the differentiation process in the GFP/CCl4 model. We performed immunohistochemical analysis to identify an important growth factor in our model, viz., fibroblast growth factor (FGF). In liver samples, the expression of FGF1 and FGF2 and of FGF receptors (FGFRs; FGFR1, FGFR2) was significantly elevated with time after bone marrow transplantation (BMT) compared with other factors, and co-expression of GFP and FGFs or FGFRs could be detected. We then analyzed the effect and molecular mechanism of FGF signaling on the enhancement of BMC differentiation into hepatocytes by immunohistochemistry, immunoblotting, and microarray analysis. Treatment with recombinant FGF (rFGF), especially rFGF2, elevated the repopulation rate of GFP-positive cells in the liver and significantly increased the expression of both Liv2 (hepatoblast marker) and albumin (hepatocyte marker). Administration of rFGF2 at BMT also raised serum albumin levels and improved the survival rate. Transplantation of BMCs with rFGF2 specifically activated tumor necrosis factor-alpha (TNF-α) signaling. Thus, FGF2 facilitates the differentiation of transplanted BMCs into albumin-producing hepatocytes via Liv2-positive hepatoblast intermediates through the activation of TNF-α signaling. Administration of FGF2 in combination with BMT improves the liver function and prognosis of mice with CCl4-induced liver damage.

Identification of common or distinct genes related to antitumor activities of a medicinal herb and its major component by oligonucleotide microarray

Although the physiological actions of many herbs are gradually being elucidated at the molecular level, it remains unclear how individual components of herbs contribute to their biological activities. In the present study, the antiproliferative activity of Coptidis rhizoma, a medicinal herb, and the major component berberine was investigated in 8 human pancreatic cancer cell lines. Gene expression patterns associated with sensitivities to each agent were analyzed with oligonucleotide arrays that comprised approximately 11,000 genes. We used a tetrazolium dye (MTT) assay to determine ID50 values after the 8 cell lines were exposed to the 2 agents for 72 hr. The ID50 value for berberine was correlated positively with that for C. rhizoma (r=0.725, p=0.0401). C. rhizoma killed tumor cells more effectively than purified berberine when normalized to the level of berberine present in the herb. From the oligonucleotide array data, we selected 20 and 13 genes with strong correlations (r2>0.81) to ID50 values for berberine and C. rhizoma, respectively. Among these 33 genes, the levels of expression of 12 were correlated with the ID50 values of both agents, suggesting that these genes are associated with tumor‐killing activity of berberine in C. rhizoma. Expression of the remaining 21 genes was correlated with the ID50 value of either purified berberine or C. rhizoma. Thus, we identified common and distinct genes responsible for anti‐proliferative activities of purified berberine and C. rhizoma. This strategy may improve our understanding of the actions of herbs with antitumor activities.

Tumor HLA-DR expression linked to early intrahepatic recurrence of hepatocellular carcinoma

The outcome of patients with hepatocellular carcinoma (HCC) remains poor because of the high frequency of intrahepatic recurrence (IHR), particularly early IHR within 1 year of hepatectomy. To search for genes involved in early IHR, we performed DNA microarray analysis in a training set of 33 HCCs and selected 46 genes linked to early IHR from approximately 6,000 genes by means of a supervised learning method. Gene selection was validated by a false discovery rate of 0.37%. The 46 genes included many immune response‐related genes, which were all downregulated in HCCs with early IHR. Four of these genes (HLA‐DRA, HLA‐DRB1, HLA‐DG and HLA‐DQA), encoding MHC class II antigens, were coordinately downregulated in HCCs with early IHR compared to levels in HCCs with nonrecurrence. A cluster analysis reproduced expression patterns of the 4 MHC class II genes in 27 blinded HCC samples. To localize the major site of production of HLA‐DR protein in the tumor, we used 50 frozen specimens from 50 HCCs. Immunofluorescence staining showed that HLA‐DR protein levels in tumor cells, but not in stromal cells, were associated with the transcription levels of HLA‐DRA determined by both DNA microarray analysis and real‐time quantitative reverse transcription‐PCR. Univariate analysis showed that tumor HLA‐DR protein expression, pTNM stage and venous invasion were associated with early IHR. Multivariate analysis showed that tumor HLA‐DR protein expression was one of the independent risk factors for early IHR, suggesting HLA‐DR protein potential as a biomarker and a molecular target for therapeutic intervention.

Gene expression profile linked to p53 status in hepatitis C virus‐related hepatocellular carcinoma

To clarify the role of p53 in 22 hepatitis C virus (HCV)‐infected hepatocellular carcinomas (HCCs), we compared the gene expression profiles of HCCs with wild‐type p53 (wt‐p53) (n=17) and those with mutant‐type p53 (mt‐p53) (n=5) by oligonucleotide microarray analysis. Among 83 p53‐related genes identified by a supervised learning method, 25 were underexpressed, and 58 were overexpressed in mt‐p53 HCCs compared with wt‐p53 HCCs. With a computer search, we identified consensus p53‐binding sequences in the 3‐kb region upstream of the translation initiation site in 59 of the 83 genes, suggesting that the in vivo p53‐associated transcription system is very complicated. These data will provide additional insights into p53‐related pathogenesis in HCV‐infected HCC.

Recognition of handwritten numerals using Gabor features

We study a Gabor filter-based feature extraction method for handwritten numeral character recognition. The performance of the Gabor filter-based method is demonstrated on the ETL-1 database. Experimental results suggest that the Gabor filter-based method should be considered in recognition of handwritten numeric characters.

On the behavior of artificial neural network classifiers in high-dimensional spaces

It is widely believed in the pattern recognition field that when a fixed number of training samples is used to design a classifier, the generalization error of the classifier tends to increase as the number of features gets larger. In this paper, we discuss the generalization error of the artificial neural network (ANN) classifiers in high-dimensional spaces, under a practical condition that the ratio of the training sample size to the dimensionality is small. Experimental results show that the generalization error of ANN classifiers seems much less sensitive to the feature size than 1-NN, Parzen and quadratic classifiers.