Shigeyuki Oba

Novel risk stratification of patients with neuroblastoma by genomic signature, which is independent of molecular signature

Human neuroblastoma remains enigmatic because it often shows spontaneous regression and aggressive growth. The prognosis of advanced stage of sporadic neuroblastomas is still poor. Here, we investigated whether genomic and molecular signatures could categorize new therapeutic risk groups in primary neuroblastomas. We conducted microarray-based comparative genomic hybridization (array-CGH) with a DNA chip carrying 2464 BAC clones to examine genomic aberrations of 236 neuroblastomas and used in-house cDNA microarrays for gene-expression profiling. Array-CGH demonstrated three major genomic groups of chromosomal aberrations: silent (GGS), partial gains and/or losses (GGP) and whole gains and/or losses (GGW), which well corresponded with the patterns of chromosome 17 abnormalities. They were further classified into subgroups with different outcomes. In 112 sporadic neuroblastomas, MYCN amplification was frequent in GGS (22%) and GGP (53%) and caused serious outcomes in patients. Sporadic tumors with a single copy of MYCN showed the 5-year cumulative survival rates of 89% in GGS, 53% in GGP and 85% in GGW. Molecular signatures also segregated patients into the favorable and unfavorable prognosis groups (P=0.001). Both univariate and multivariate analyses revealed that genomic and molecular signatures were mutually independent, powerful prognostic indicators. Thus, combined genomic and molecular signatures may categorize novel risk groups and confer new clues for allowing tailored or even individualized medicine to patients with neuroblastoma.

Identification of expressed genes linked to malignancy of human colorectal carcinoma by parametric clustering of quantitative expression data

BackgroundIndividual human carcinomas have distinct biological and clinical properties: gene-expression profiling is expected to unveil the underlying molecular features. Particular interest has been focused on potential diagnostic and therapeutic applications. Solid tumors, such as colorectal carcinoma, present additional obstacles for experimental and data analysis.ResultsWe analyzed the expression levels of 1,536 genes in 100 colorectal cancer and 11 normal tissues using adaptor-tagged competitive PCR, a high-throughput reverse transcription-PCR technique. A parametric clustering method using the Gaussian mixture model and the Bayes inference revealed three groups of expressed genes. Two contained large numbers of genes. One of these groups correlated well with both the differences between tumor and normal tissues and the presence or absence of distant metastasis, whereas the other correlated only with the tumor/normal difference. The third group comprised a small number of genes. Approximately half showed an identical expression pattern, and cancer tissues were classified into two groups by their expression levels. The high-expression group had strong correlation with distant metastasis, and a poorer survival rate than the low-expression group, indicating possible clinical applications of these genes. In addition to c-yes, a homolog of a viral oncogene, prognostic indicators included genes specific to glial cells, which gives a new link between malignancy and ectopic gene expression.ConclusionsThe malignancy of human colorectal carcinoma is correlated with a unique expression pattern of a specific group of genes, allowing the classification of tumor tissues into two clinically distinct groups.

Molecular features of non-B, non-C hepatocellular carcinoma: a PCR-array gene expression profiling study

BACKGROUND/AIMS Hepatocellular carcinoma (HCC) usually develops following chronic liver inflammation caused by hepatitis C or B virus. Through expression profiling in a rare type of HCC, for which the causes are unknown, we sought to find key genes responsible for each step of hepatocarcinogenesis in the absence of viral influence. METHODS We used 68 non-B, non-C liver tissues (20 HCC, 17 non-tumor, 31 normal liver) for expression profiling with PCR-array carrying 3072 genes known to be expressed in liver tissues. To select the differentially expressed genes, we performed random permutation testing. A weighted voting classification algorithm was used to confirm the reliability of gene selection. We then compared these genes with the results of previous expression profiling studies. RESULTS A total of 220 differentially expressed genes were selected by random permutation tests. The classification accuracies using these genes were 91.8, 92.0 and 100.0% by a leave-one-out cross-validation, an additional PCR-array dataset and a Stanford DNA microarray dataset, respectively. By comparing our results with previous reports on virus-infected HCC, four genes (ALB, A2M, ECHS1 and IGFBP3) were commonly selected in some studies. CONCLUSIONS The 220 differentially expressed genes selected by PCR-array are potentially responsible for hepatocarcinogenesis in the absence of viral influence.

Prediction of recurrence in advanced gastric cancer patients after curative resection by gene expression profiling

The prognosis of patients with advanced gastric cancer remains unfavorable. Even after curative resection, 40% of patients with advanced gastric cancer die of recurrence. Conventional clinicopathlogic findings are sometimes inadequate for predicting recurrence in individuals. Hence, we tried to construct a new diagnostic system, which predicts recurrence in patients with advanced gastric cancer after curative resection based on molecular analysis. Gastric cancer progression is a function of multiple genetic events that may affect the expression of large number of genes. We performed gene expression profiling with 2,304 genes in 60 advanced gastric cancer patients who underwent curative resection using a PCR array technique, a high‐throughput quantitative RT‐PCR technique. The diagnostic system, which was constructed from the learning set comprised of 40 patients with the most informative 29 genes, classified each case into a good‐signature or poor‐signature group. Then, we confirmed the predictive performance in an additional test set comprised of 20 patients, and the prediction accuracy for recurrence was 75%. Kaplan‐Meier analysis revealed significant difference between the good‐signature and the poor‐signature group (p = 0.0125). Especially in patients with smaller tumor (≤ 5 cm), less developed LN metastasis (N0,1), or earlier stage (stages I and II), the prediction accuracy was high (88.9%, 84.6%, or 81.8%, respectively). Our diagnostic system based on systematic analysis of gene expression profiling can predict the recurrence at clinically meaningful level. By combining our system with conventional clinicopathologic factors, we can improve the prediction of recurrence in patients with advanced gastric cancer who underwent curative surgery.

A multi-class predictor based on a probabilistic model: application to gene expression profiling-based diagnosis of thyroid tumors

BackgroundAlthough microscopic diagnosis has been playing the decisive role in cancer diagnostics, there have been cases in which it does not satisfy the clinical need. Differential diagnosis of malignant and benign thyroid tissues is one such case, and supplementary diagnosis such as that by gene expression profile is expected.ResultsWith four thyroid tissue types, i.e., papillary carcinoma, follicular carcinoma, follicular adenoma, and normal thyroid, we performed gene expression profiling with adaptor-tagged competitive PCR, a high-throughput RT-PCR technique. For differential diagnosis, we applied a novel multi-class predictor, introducing probabilistic outputs. Multi-class predictors were constructed using various combinations of binary classifiers. The learning set included 119 samples, and the predictors were evaluated by strict leave-one-out cross validation. Trials included classical combinations, i.e., one-to-one, one-to-the-rest, but the predictor using more combination exhibited the better prediction accuracy. This characteristic was consistent with other gene expression data sets. The performance of the selected predictor was then tested with an independent set consisting of 49 samples. The resulting test prediction accuracy was 85.7%.ConclusionMolecular diagnosis of thyroid tissues is feasible by gene expression profiling, and the current level is promising towards the automatic diagnostic tool to complement the present medical procedures. A multi-class predictor with an exhaustive combination of binary classifiers could achieve a higher prediction accuracy than those with classical combinations and other predictors such as multi-class SVM. The probabilistic outputs of the predictor offer more detailed information for each sample, which enables visualization of each sample in low-dimensional classification spaces. These new concepts should help to improve the multi-class classification including that of cancer tissues.

Optimal Aggregation of Binary Classifiers for Multiclass Cancer Diagnosis Using Gene Expression Profiles

Multiclass classification is one of the fundamental tasks in bioinformatics and typically arises in cancer diagnosis studies by gene expression profiling. There have been many studies of aggregating binary classifiers to construct a multiclass classifier based on one-versus-the-rest (1R), one-versus-one (11), or other coding strategies, as well as some comparison studies between them. However, the studies found that the best coding depends on each situation. Therefore, a new problem, which we call the “optimal coding problem,” has arisen: how can we determine which coding is the optimal one in each situation? To approach this optimal coding problem, we propose a novel framework for constructing a multiclass classifier, in which each binary classifier to be aggregated has a weight value to be optimally tuned based on the observed data. Although there is no a priori answer to the optimal coding problem, our weight tuning method can be a consistent answer to the problem. We apply this method to various classification problems including a synthesized data set and some cancer diagnosis data sets from gene expression profiling. The results demonstrate that, in most situations, our method can improve classification accuracy over simple voting heuristics and is better than or comparable to state-of-the-art multiclass predictors.

Numerical indices based on circulating tumor DNA for the evaluation of therapeutic response and disease progression in lung cancer patients.

Monitoring of disease/therapeutic conditions is an important application of circulating tumor DNA (ctDNA). We devised numerical indices, based on ctDNA dynamics, for therapeutic response and disease progression. 52 lung cancer patients subjected to the EGFR-TKI treatment were prospectively collected, and ctDNA levels represented by the activating and T790M mutations were measured using deep sequencing. Typically, ctDNA levels decreased sharply upon initiation of EGFR-TKI, however this did not occur in progressive disease (PD) cases. All 3 PD cases at initiation of EGFR-TKI were separated from other 27 cases in a two-dimensional space generated by the ratio of the ctDNA levels before and after therapy initiation (mutation allele ratio in therapy, MART) and the average ctDNA level. For responses to various agents after disease progression, PD/stable disease cases were separated from partial response cases using MART (accuracy, 94.7%; 95% CI, 73.5–100). For disease progression, the initiation of ctDNA elevation (initial positive point) was compared with the onset of objective disease progression. In 11 out of 28 eligible patients, both occurred within ±100 day range, suggesting a detection of the same change in disease condition. Our numerical indices have potential applicability in clinical practice, pending confirmation with designed prospective studies.

Prior hyperparameters in Bayesian PCA

Bayesian PCA (BPCA) provides a Bayes inference for probabilistic PCA, in which several prior distributions have been devised; for example, automatic relevance determination (ARD) is used for determining the dimensionality. However, there is arbitrariness in prior setting; different prior settings result in different estimations. This article aims at presenting a standard setting of prior distribution for BPCA. We first define a general hierarchical prior for BPCA and show an exact predictive distribution. We show that several of the already proposed priors can be regarded as special cases of the general prior. By comparing various priors, we show that BPCA with nearly non-informative hierarchical priors exhibits the best performance.

Variational Bayes method for Mixture of Principal Component Analyzers

When dealing with high-dimensional data, it is proper to construct an appropriate feature space consisting of relevant features extracted from given data. Mixture of Principal Component Analyzers (MPCA) is a feature extracting tool in which clustering in the data space and principal component analysis in each cluster are simultaneously done. A probabilistic model of MPCA and its maximum likelihood (ML) inference were presented by Tipping and Bishop (1999). Although the probabilistic formulation can be satisfactorily advantageous within Bayesian inference instead of ML inference, Bayesian inference often requires difficult integrations. Such difficulty comes to be overcome by the variational Bayes (VB) approximation that has recently been developed. In this paper, we present a VB inference algorithm for MPCA, and its application to a recognition problem of handwritten digit images.

Missing Value Estimation Using Mixture of PCAs

We apply mixture of principal component analyzers (MPCA) to missing value estimation problems. A variational Bayes (VB) method for MPCA with missing values is developed. The missing values are regarded as hidden variables aud their estimation is done simultaneously with the parameter estimation. It is found that VB method is better than maximum likelihood method by using artificial data. We also applied our method to DNA microarray data and the performance outweighed the conventional k-nearest neighbor method.