New classification of digestive system cancers based on gene expression analysis.

利用cDNA筛选阵列对1300个与癌症和免疫相关的基因进行筛选，对结肠癌和直肠癌总RNA进行基因表达分析。样本数量为51个，其中6个样本重叠。本研究旨在建立一种新的分子水平预后方法。我们研究了以下两点：1)通过构建基于基因表达谱的肝转移预测模型建立预后方法；2)提取一组肝转移相关基因。我们选择了神经网络方法构建分类器，并采用遗传算法对其进行优化，并在此基础上与合作伙伴MediBIC有限公司共同开发了该系统。(东京)。通过使用发表在《自然医学》上的公开数据验证了我们系统的准确性。结果，平均正确率为97.8%。这些结果表明我们的系统和模型更适合于预测肿瘤转移的可能性。然后，利用临床数据，将该系统应用于肝转移预测模型的构建。在我们的病例中，对于肝转移，该系统的准确性也达到了足够的水平。不幸的是，以前报道的与肝转移有关的主要分子在基于我们的模型提取的基因列表中没有发现。然而，通过进一步研究提取的基因，仍有可能找到新的生物标志物。由于我们认识到总RNA样品制备的问题，因此我们保留了对生物标记候选物是否包括在基因列表中的研究。在本研究中，我们的重点是鉴别是否发生肝转移的分类器。然而，我们意识到肝转移应该分为两组，一组是同步转移，另一组是异时转移。我们开始了一系列的研究，采用三重分类方法，即同步肝转移（5个样本），异时性肝转移（5个样本）和无肝转移（5个样本）。目前，我们正在检查将用于该系列研究的所有样品的质量。少

Gene expression analysis with total RNA extracted from colon and rectal cancer was performed by using of cDNA filter arrays spotted with 1,300 genes relating to cancer and immunity. The number of samples was 51 including six samples overlapped. This study was aimed at the establishment of a new molecular-level prognosis method. We investigated following two points, that is, 1) establishment of prognosis method by constructing the liver metastasis predicting model based on gene expression profiling and 2) extraction of a group of genes relating to liver metastasis. We had chosen a neural network method to build a classifier and a genetic algorithm for its optimization, and then developed the system based on the method with our partner, MediBIC Co.Ltd. (Tokyo). The accuracy of our system was verified by using of the public data, published in Nature Medicine. As a result, an average rate of correct answer was 97.8%. These results suggest that our system and model were preferable for predi … More cting the possibility of metastasis. Then, we applied the system to the construction of liver metastasis forecasting model by using of clinical data. The accuracy of the system was sufficient level also in our case regarding the liver metastasis. Major molecules previously reported to relate to the liver metastases had not unfortunately found in the list of genes extracted based on our model. However, there is still a possibility to find a novel biomarker by investigating the extracted genes further. Because we recognized problems on the total RNA sample preparation, we have left it open to be investigated whether the biomarker candidates are included in the gene list. In this research, we were focusing on the classifier discriminating whether the liver metastasis occurs or not. We are aware however that liver metastasis should to be analyzed by separating into two groups, one is synchronous metastasis and the other is metachronous metastasis. We have started a series of study taking triple classifier approaches, namely, synchronous liver metastasis (5samples), metachronous liver metastasis (5samples) and liver metastasis free (5samples). We are checking the quality of all samples that will be used in the series of the study, at present. Less