Rescaled range R/S analysis application for genes prediction in the plant genome

摘要

Currently gene's prediction problem is one of the main genomic challenges. Prediction allows performing experiments with high probability of interesting genes to be found and compare DNA regions of agronomic importance among genomes; besides, it helps to restrict the searching spaces into the data bases. A statistical procedure based on the R/S analysis and the Hurst coefficient was developed in order to characterize and predict genes and their structural components (exones and intrones) in the whole eukaryotic genomes of Arabidopsis thaliana, Oriza sativa and Mus musculus. Python programming language algorithms were developed with the purpose of extract, screen and modeling more than 80% of the registered gene sequences for these genomes in the NCBI Gene Bank data base. The R/S analysis allows to demonstrate that a structural order do exist in the distribution of the nucleotides which are constituting sequences with the memory or long range dependence phenomena predominance. The memory structure varies according to the sequences type and the species genome. The genes and exones sequences from the analyzed plant genomes showed a persistent behavior whereas those from the intrones had an anti-persistent behavior, in comparison with animal genome in which the three type of sequences showed persistent behavior. According to R/S analysis out coming parameters the genome sequences distribution pattern was replicated in a statistically similar manner in each chromosome belonging to one species, constituting fundamental evidences of invariance by scale change; it means each chromosome by itself is a statistical replication to a minor scale of the whole genome. The parameters constituted compact criteria in order to derivate sequences predictors (classifiers) which reached sensibility and specificity averages higher than 81% and 70% respectively. This procedure could be tried in other genomes and be used as a criterion in order to increasing selection efficiency in plant genetic breeding programs.