Modelling of gene signal attribute reduction based on neighbourhood granulation and rough approximation

The update of high-throughput sequencing technology has led to the dramatic increase in the number of sequenced meta-genomic DNA sequences. However, extracting a nearly 10,000-dimensional digital signature as a species tag will inevitably bring about tremendous computational load. Therefore, how to reduce the macro features of macro-genomic DNA, how to extract and select the subset with the best characteristics as a species tag, has become a research direction of bio-informatics. In this paper, we use neighbourhood granulation and rough approximation theory modelling to study the method of attribute reduction of meta-genomic DNA fragments and to deduce the digital features of meta-genome at the 'genus' classification. The results show that this method can be effective to screen out representative species tags and improve classification efficiency.