Knowledge-based prediction of DNA hydration using hydrated dinucleotides as building blocks

DNA hydration is a key factor in stabilizing DNA structure and influencing its interactions. Prior research has shown that water molecules form distinct density patterns around DNA. However, the precise relationship between DNA sequence, structure, and hydration remains unclear. Existing studies have focused on general hydration patterns without accounting for sequence-specific effects. This gap motivated the development of methods that integrate both sequence and structural information to model hydration. No prior work had resolved how dinucleotide sequences and conformers influence hydration site positions. This paper introduces a novel approach using hydrated dinucleotides as building blocks. The study builds on prior knowledge of DNA hydration but extends it by incorporating sequence and structural variability. The authors aim to bridge the gap between hydration modeling and experimental validation.

The study aimed to develop a predictive model of DNA hydration using hydrated dinucleotides as structural units. The authors sought to understand how sequence and structure jointly determine hydration site positions. By analyzing a large dataset of DNA structures, the researchers aimed to classify dinucleotides into hydration categories. The goal was to create a framework for predicting hydration in new DNA structures. The study also aimed to validate the model against independent experimental data. The researchers wanted to assess whether hydration patterns could be reliably predicted using dinucleotide building blocks. The ultimate purpose was to provide a tool for interpreting experimental DNA hydration data. The study aimed to demonstrate the utility of hydrated building blocks in both computational and experimental contexts.

The researchers analyzed 2727 nonredundant DNA chains containing 41,853 dinucleotides and 316,265 associated water molecules. Dinucleotides were categorized by sequence and structural conformers known as nucleotide conformers (NtCs). Hydration sites were identified as peaks in water density distributions around each dinucleotide. The study used probability-based calculations to model hydration site positions. The dataset included crystal and NMR structures to ensure structural diversity. The researchers developed a classification system based on sequence and structural features. The hydration model was tested on an independent set of ten DNA structures. Predicted hydration sites were compared to experimentally observed water positions to assess accuracy.

The study identified hydration sites (HSs) as peaks in water density distributions around dinucleotides. Predicted HSs matched experimental water positions in ten structures with over 40% within 0.5 Å. The model successfully reproduced known hydration features like the `spine of hydration' in B-DNA. The method captured sequence-specific hydration patterns across different dinucleotide sequences. Structural conformers influenced hydration site positions more than sequence alone in some cases. The hydration model showed strong agreement with experimental data in crystal and NMR structures. The study demonstrated that hydration is sequence- and structure-dependent at the dinucleotide level. The data and predictions are publicly available at https://watlas.datmos.org/watna/.

The authors propose that hydrated dinucleotides can serve as predictive building blocks for DNA hydration modeling. The results suggest that sequence and structure jointly determine hydration site positions. The model accurately reproduces known hydration features in B-DNA and other structures. The study demonstrates that hydration patterns can be predicted using dinucleotide-level data. The hydration model aligns well with experimental hydration data from crystal and NMR structures. The authors suggest that this approach can guide the interpretation of experimental DNA hydration data. The study supports the use of hydrated building blocks in both computational and experimental contexts. The data and predictions are available for further analysis and validation.