Scoring Function And Docking Algorithm For Nucleic Acid–ligand Interactions

Nucleic acid is one of the basic components in cells,and its basic unit is nucleotide.As a carrier of genetic information,nucleic acids perform a key role in the process of information storage and transmission.Nucleic acids also play a vital role in the production,development,and reproduction of organisms.The research on nucleic acid molecules is now an important part of life science research.There is a class of nucleotides that can be used as drug targets,such as riboswitches which could combine with ligands to regulate gene expression and activate/inactivate nucleic acid biological functions.Consequently,it is necessary to study nucleic acid–ligand interactions to gain the deep understanding of nucleic acid functions.Structure determines function.However,due to the expensive costs and technical difficulty of experimental methods,the number of experimentally determined nucleic acid–ligand complex structures is still limited.However,computational methods such as molecular docking have attracted widespread attention due to their high computational efficiency and low costs.Molecular docking has played a complementary role in the complex structure modeling of nucleic acid–ligand interactions.In molecular docking,the scoring function guides the generation and screening of complex structures by evaluating the tightness and specificity of the binding between ligands and nucleic acids.Therefore,the accuracy of the scoring function will directly affect the quality of modeling results.Based on the inverse Boltzmann correlation,we trained a knowledge-based scoring functions,referred as ITScore-NL,for nucleic acid–small molecule interaction energy calculation.In the construction of the scoring function,we explicitly consider the stacking and electrostatic interactions between two molecules.Stacking is one of the main features in nucleic acid structures and plays a key role in stabling the structure of nucleic acids.Within a certain distance,when the nucleic acid base and the aromatic ring of the ligand are located in two approximately parallel planes,a stacking effect would occur.In addition,due to the highly charged nature of nucleic acids,electrostatic interactions are especially important for the binding affinity between nucleic acids and ligands.In consequence,we explicitly consider the electrostatic interactions in the derivation of scoring function to reduce the statistical errors due to insufficient frequencies of the atom pair at some distances.It could be seen from the results that the stacking potential and electrostatic interaction have different contributions to the performances of the scoring function.Namely,including the stacking term mainly increased the improvement in binding affinity prediction and obtained a good correlation coefficient.In contrast,including the electrostatic term mainly improved the power of ITScore in binding mode predictions.Under the loose and strict standards,the screening of the structure has been greatly improved.Inspired by the outstanding performance of our newly trained scoring function(ITScoreNL)in guiding the modeling of complex structure,we developed a new RNA/DNA–small molecule docking algorithm,called NLDock.The new docking program is a major modification of the protein–ligand docking program MDock.The docking program simulates the ligand flexibility by docking multiple ligand conformations.The results show that NLDock is significantly better than the other three docking programs in terms of predicting the binding mode of local rigid ligand docking and flexible ligand docking.The good performance of NLDock may be due to the advantages of ITScore-NL scoring function in RNA/DNA–small molecule binding affinity prediction.At the same time,in terms of global docking where the binding site is unknown,the possible binding regions of small molecules are predicted based on geometric criterias.Then a local docking is performed,so as to achieve the purpose of global docking.The docking results show that,compared with the other three docking algorithms,NLDock achieved more competitive results.In addition,the compuatational efficiency of docking is also a major advantage of NLDock.The average docking time of NLDock is nearly an order of magnitude faster than the other docking software.In molecular docking,when the binding site is unknown,identifying the possible binding regions of candidate molecules is needed.Usually,the binding region is determined based on personal experience or information provided by experiments.This is more difficult for researches who are new to this field.Blind docking has always been a more challenging task,and it is also the hinder in improving the precision of molecular docking.In order to accurately predict nucleic acid–ligand binding sites,we develop a binding site predicted algorithm,called NLsite.NLsite is a novel method that integrates geometry and energy function to predict the RNA binding sites.To start,the RNA is configured in a threedimensional grid,and the grid points located in the concave area of RNA structure are selected;Moreover,a number of balls with a given atom type are placed in the concave area.During this process,the energy of each grid point in those regions is calculated based on ITScore-NL.The grid points are filtrated again according to their energy.Finally,the grid point energy is projected onto its surrounding nucleotides by distance mapping.Given a certain energy threshold,nucleotides within the energy threshold are considered as the nucleotides near the binding site.The results show that NLsite is an effective site prediction method.After the validation on five test sets,the results are compared with other three algorithms.The four criterions are used to show the precision and robustness of NLsite.