Kullback-Leibler cluster entropy to quantify local correlation in human genes

In this study, we apply an information theoretical clustering method based on the Kullback-Leibler relative entropy, to analyse the structure of the human gene. The gene sequence is first converted into numerical representation, then its divergence from synthetic fractional Brownian motion models taken as a reference is quantified. The functional D_C[P∥Q] is estimated using the probability distribution P obtained by ranking the clusters generated in the nucleotide sequences of the gene and compared against the distribution Q drawn from synthetic data sequences ranging from fully uncorrelated to partially correlated fractional Brownian motions adopted as reference models. The biological relevance of the Kullback–Leibler divergence is scrutinized by analysing the local correlation correlation with different structures within the gene: exons, the enhancers, and epigenetically modified accessible regions (EMARs). Our results show that the observed patterns remain consistent across different segmentation strategies, including both overlapping and non-overlapping windows, and highlight the potential of this method for broader applications in comparative genomics and gene function annotation.