Sequence alignment-free method for phylogenetic and functional prediction and its application to molecular evolutionary studies

碱基和氨基酸序列的同源性搜索已成为生物信息学领域的一个基本工具，它不仅对分析基因和蛋白质的进化和同源性至关重要，而且对估计已测序基因组中每个蛋白质基因的功能也至关重要。虽然序列同源性搜索显然是有用的，但这种方法不能估计新测序的基因组中几乎一半基因的蛋白质功能，特别是新物种。为了补充序列同源性搜索，迫切需要建立基于不同原理的蛋白质功能估计方法。将BLSOM（Batch-learning SOM）方法应用于蛋白质氨基酸序列，建立了一种基于寡肽频率相似性的聚类函数估计方法。用于寡核苷酸频率的BL-SOM可以识别个体基因组中寡核苷酸频率的基因组特异性特征，允许根据物种对基因组片段序列进行分类（自组织），而不需要在计算期间提供物种信息。我们将BLSOMs应用于宏基因组分析获得的大量环境基因组序列的系统发育和功能预测。

The homology search of base and amino acid sequences has become a basic tool in the bioinformatics field, which is essential not only for analyzing the evolution and phylogeny of genes and proteins but also for estimating the function of each protein gene in sequenced genomes. While sequence homology search is obviously useful, this method cannot estimate protein functions for almost half of genes in newly sequenced genomes, especially of novel species. To complement the sequence homology search, it is urgently required to establish methods for estimating protein functions based on different principles. By applying the BLSOM (Batch-learning SOM) method to amino-acid sequences of proteins, we established a function estimation method that utilizes clustering based on the similarity of oligopeptides frequencies. The BL-SOM for oligonucleotide frequencies could recognize genome-specific characteristics of oligonucleotide frequencies in individual genomes, permitting classification (self-organization) of genomic fragment sequences according to species without the need for species information during the calculation. We applied the BLSOMs to phylogenetic and functional prediction of a massive amount of environmental genomic sequences obtained by metagenomic analyses.