(m, n)-mer-a simple statistical feature for sequence classification.

ABSTRACT:

Summary

The (m, n)-mer is a simple alternative classification feature based on conditional probability distributions. In this application note, we compared k-mer and (m, n)-mer frequency features in 11 distinct datasets used for binary, multiclass and clustering classifications. Our findings show that the (m, n)-mer frequency features are related to the highest performance metrics and often statistically outperformed the k-mers. Here, the (m, n)-mer frequencies improved performance for classifying smaller sequence lengths (as short as 300 bp) and yielded higher metrics when using short values of k (ranging from 2 to 4). Therefore, we present the (m, n)-mers frequencies to the scientific commu

SUBMITTER: de Andrade AAS 

PROVIDER: S-EPMC10338135 | biostudies-literature | 2023

REPOSITORIES: biostudies-literature