An effective SNR gene prediction algorithm based on merge of nucleotide segments (MNS)

Signal processing approaches, among the deterministic approaches in gene prediction, have been attracting significant attentions in genomic DNA research for its fine model-independent feature and less reliance on known datasets. An effective SNR gene prediction algorithm based on merge of nucleotide segments (MNS) was proposed in this paper. A fast calculation equation for SNR was also derived. The new algorithm is effective, efficient and could be applied to various higher life forms, especially mammals. The AUC of MNS on Homo sapiens & mus musculus is 0.8073, and the AUC of MNS on various mammals is 0.7780, which are satisfying results among deterministic approaches. It is anticipated that the novel algorithm is promising for the prediction of various higher life forms, of which the DNA data information are limited. Testing results of MNS on various species implied the relationship between a species´ period-3 property and its evolutionary scale, which indicates that MNS could also be utilized as an auxiliary approach for taxonomy.