Profound Non-Randomness in Dinucleotide Arrangements within Ultra-Conserved Non-Coding Elements and the Human Genome

Abstract

Long human ultra-conserved non-coding elements (UCNEs) do not have any sequence similarity to each other or other characteristics that make them unalterable during vertebrate evolution. We hypothesized that UCNEs have unique dinucleotide (DN) composition and arrangements compared to the rest of the genome. A total of 4272 human UCNE sequences were analyzed computationally and compared with the whole genomes of human, chicken, zebrafish, and fly. Statistical analysis was performed to assess the non-randomness in DN spacing arrangements within the entire human genome and within UCNEs. Significant non-randomness in DN spacing arrangements was observed in the entire human genome. Additionally, UCNEs exhibited distinct patterns in DN arrangements compared to the rest of the genome. Approximately 83% of all DN pairs within UCNEs showed significant (>10%) non-random genomic arrangements at short distances (2–6 nucleotides) relative to each other. At the extremes, non-randomness in DN spacing distances deviated up to 40% from expected values and were frequently associated with GpC, CpG, ApT, and GpG/CpC dinucleotides. The described peculiarities in DN arrangements have persisted for hundreds of millions of years in vertebrates. These distinctive patterns may suggest that UCNEs have specific DNA conformations.

Computational identification of ultra-conserved elements in the human genome: a hypothesis on homologous DNA pairing

“…One of the prominent patterns of dinucleotide arrangements described in Fedorova et al. ( 16 ) is shown in Figure 3 . Seventeen different dinucleotide pairs have the same peak across UCNE curves.…”

“… Distribution of spacing distances between pairs of particular dinucleotides for UCNE (red), uWGE (blue), random UCNE (yellow) and random uWGE (gray) as described in Fedorova et al. ( 16 ). The 99.7% confidence intervals (±3σ) are demonstrated for UCNE datasets as vertical bars.…”

“…For distances between dinucleotides, we used the measurement scheme from our previous paper by Fedorova et al. ( 16 ). The minimal distance corresponds to the shortest distance between intersected dinucleotides.…”

“…Recently, our team demonstrated that the only common nucleotide similarity across UCNE sequences is their unique dinucleotide composition ( 12 , 16 ). In general, UCNEs are GC-poor nucleotide sequences that are strongly enriched with GpC dinucleotides and deficient in GpG and CpC dinucleotides.…”

DNA N-gram Analysis Framework (DNAnamer): A generalized N-gram frequency analysis framework for the supervised classification of DNA sequences

“…H2 requires that DNA N-grams such as bi-, tri-, and tetra-nucleotide pairings do not occur randomly, and H3 requires that DNA N-gram frequency patterns are conserved and identifiable. In fact, it has been demonstrated that di-, tri-, and poly-nucleotide repeats are not randomly distributed and are highly conserved with complex and distinct patterns [9][10][11]. Given H2 and H3, DNA N-gram frequency patterns can be leveraged to construct a generalized stochastic model that can be used to efficiently solve a wide variety of supervised classification and data reduction problems highly relevant to genomics and genetics.…”