We report an analysis of Homo sapiens DNA through the formalism of κ statistics, which encompasses power-law correlations and provides an optimization principle that permits us to model distinct physical systems; i.e., the power-law distribution of the length of DNA bases is calculated from a general model which follows arguments similar to those proposed in Maxwell's deduction of statistical distributions. The viability of the model is tested using a data set from a catalog of proteins collected from the Ensembl Project. The results indicate that the short-range correlations, always present in coding DNA sequences, are appropriately captured through the Kaniadakis power-law distribution, adequately describing the cumulative length distribution of DNA bases, in contrast with the case of the traditional exponential statistical model.