Partitioning the Human Transcriptome Using <i>HKera</i>, a Novel Classifier of Housekeeping and Tissue-Specific Genes

High-throughput transcriptomic experiments have made it possible to classify genes that are ubiquitously expressed as housekeeping (HK) genes and those expressed only in selective tissues as tissue-specific (TS) genes. Although partitioning a transcriptome into HK and TS genes is conceptually problematic owing to the lack of precise definitions and gene expression profile criteria for the two, information whether a gene is an HK or a TS gene can provide an initial clue to its cellular and/or functional role. Consequently, the development of new and novel HK (TS) classification methods has been a topic of considerable interest in post-genomics research. Here, we report such a development. Our method, called HKera, differs from the others by utilizing a novel property of HK genes that we have previously uncovered, namely that the ranking order of their expression levels, as opposed to the expression levels themselves, tends to be preserved from one tissue to another. Evaluated against multiple benchmark sets of human HK genes, including one recently derived from second generation sequencing data, HKera was shown to perform significantly better than five other classifiers that use different methodologies. An enrichment analysis of pathway and gene ontology annotations showed that HKera-predicted HK and TS genes have distinct functional roles and, together, cover most of the ontology categories. These results show that HKera is a good transcriptome partitioner that can be used to search for, and obtain useful expression and functional information for, novel HK (TS) genes.高通量转录组实验（High-throughput transcriptomic experiments）已可将广泛表达的基因归类为持家基因（housekeeping, HK），仅在特定组织中表达的基因归类为组织特异性基因（tissue-specific, TS）。尽管由于缺乏两类基因的精确定义与基因表达谱判定标准，将转录组划分为HK与TS基因的做法在概念上存在不严谨之处，但明确某一基因属于HK还是TS，可为其细胞功能与功能作用提供初步线索。因此，开发新型HK（TS）分类方法已成为后基因组学研究中的热点课题。本文报道了此类方法的研发成果：我们提出的HKera方法，其独特优势在于利用了此前发现的HK基因新特性——即HK基因的表达水平排序（而非表达水平本身），在不同组织间往往具有保守性。以多组人类HK基因基准数据集（包含一组新近基于二代测序数据（second generation sequencing data）构建的基准集）进行评估，结果显示HKera的性能显著优于其余5种采用不同方法的分类器。通路与基因本体（gene ontology）注释富集分析表明，HKera预测得到的HK与TS基因具有截然不同的功能特征，二者共同覆盖了绝大多数基因本体分类类别。上述结果证实，HKera是一款性能优异的转录组划分工具，可用于搜寻新型HK（TS）基因并获取其有价值的表达与功能信息。