Abstract To gain a detailed understanding of protein structure, function, and interaction, water molecules around proteins are important. Therefore, computational methods for predicting water positions are required. When a hydration water distribution such as a 3D distribution function is available, methods to predict water positions explicitly from the water distribution are useful. In this paper, we introduce DroPred, a method for predicting water positions based on a 3D distribution function of water oxygen atoms using a weighted Monte Carlo method. The probability density derived from the 3D distribution function is used as weight in the weighted Monte Carlo method. DroPred generates multiple samples from a single 3D distribution function. We evaluated the performance of DroPred by predicting water positions at protein–protein interface structures. By adjusting the weight using an exponential parameter, prediction performance of DroPred in water position sampling was improved. This method will be helpful for understanding protein structure, function, and interaction.
Read full abstract