Nabe: an energetic database of amino acid mutations in protein–nucleic acid binding interfaces

Abstract

Protein–nucleic acid complexes play essential roles in regulating transcription, translation, DNA replication, repair and recombination, RNA processing and translocation. Site-directed mutagenesis has been extremely useful in understanding the principles of protein–DNA and protein–RNA interactions, and experimentally determined mutagenesis data are prerequisites for designing effective algorithms for predicting the binding affinity change upon mutation. However, a vital challenge in this area is the lack of sufficient public experimentally recognized mutation data, which leads to difficulties in developing computational prediction methods. In this article, we present Nabe, an integrated database of amino acid mutations and their effects on the binding free energy in protein–DNA and protein–RNA interactions for which binding affinities have been experimentally determined. Compared with existing databases and data sets, Nabe is the largest protein–nucleic acid mutation database, containing 2506 mutations in 473 protein–DNA and protein–RNA complexes, and of that 1751 are alanine mutations in 405 protein–nucleic acid complexes. For researchers to conveniently utilize the data, Nabe assembles protein–DNA and protein–RNA benchmark databases by adopting the data-processing procedures in the majority of models. To further facilitate users to query data, Nabe provides a searchable and graphical web page. Database URL: http://nabe.denglab.org

Using machine‐learning‐driven approaches to boost hot‐spot's knowledge

“…A few databases with available experimental information are listed in Table 2, some of which gather information from other mutagenesis experiences besides alanine. For protein–protein complexes there are four main databases: the alanine scanning energetics database (ASEdb), 11 protein–protein complex mutation thermodynamics (PROXiMATE, 174 previously known as PINT 172 ), the binding interface database (BID 116 ), and structural database of kinetics and energetics of mutant protein interactions (SKEMPI), whereas for protein–nucleic acid, we can access protein–nucleic acid interactions (PRONIT 176 ) and protein–nucleic acid binding energetic database (NABE 177 ). Table 2 also includes some other curated, nonredundant datasets of mutations that satisfy a few requirements: An existing set of relative binding free energy difference (ΔΔ G binding ) values for interfacial residues coming from experimental alanine mutagenesis; Availability of a three‐dimensional (3D)‐structure in the protein databank (PDB); and A maximum of 35% sequence identity in each interface, hence preventing repeated complexes. …”

“…A few databases with available experimental information are listed in Table 2, some of which gather information from other mutagenesis experiences besides alanine. For protein-protein complexes there are four main databases: the alanine scanning energetics database (ASEdb), 11 protein-protein complex mutation thermodynamics (PROXiMATE, 174 previously known as PINT 172 ), the binding interface database (BID 116 ), and structural database of kinetics and energetics of mutant protein interactions (SKEMPI), whereas for protein-nucleic acid, we can access protein-nucleic acid interactions (PRONIT 176 ) and proteinnucleic acid binding energetic database (NABE 177 ). 1.…”

Research on the Computational Prediction of Essential Genes

“…Experiments conducted in vitro may not truly reflect the activity of proteins in the body. Moreover, the datasets in biological databases have problems such as noise, deviation, and missing data ( Su et al, 2019 ; Liu et al, 2020d ; Liu et al, 2021b ; Su et al, 2021 ).…”

Prediction of hot spots in protein–DNA binding interfaces based on discrete wavelet transform and wavelet packet transform

“…In this work, we proposed a novel method based on discrete wavelet transform (DWT) and wavelet packet transform (WPT) to describe conventional features, termed WTL-PDH, to predict hot spots in protein–DNA binding interfaces. We screened 339 mutations in 117 protein–DNA complexes from dbAMEPNI [ 21 ], SAMPDI, Nabe [ 22 ], ProNAB [ 23 ], and then used Synthetic Minority Over-sampling Technique (SMOTE) [ 24 ] to solve the imbalance between positive and negative samples. Firstly, we extracted 43 dimensional traditional features in terms of solvent accessibility surface area, secondary structure, protrusion index and depth index, and hydrogen bond.…”