patcHwork: a user-friendly pH sensitivity analysis web server for protein sequences and structures

Abstract

pH regulates protein function and interactions by altering the charge of individual residues causing loss or gain of intramolecular noncovalent bonds, which may lead to structural rearrangements. While tools to analyze residue-specific charge distribution of proteins at a given pH exist, currently no tool is available to investigate noncovalent bond changes at two different pH values. To make protein pH sensitivity analysis more accessible, we developed patcHwork, a web server that combines the identification of amino acids undergoing a charge shift with the determination of affected noncovalent bonds at two user-defined pH values. At the sequence-only level, patcHwork applies the Henderson–Hasselbalch equation to determine pH-sensitive residues. When the 3D protein structure is available, patcHwork can be employed to gain mechanistic understanding of the effect of pH. This is achieved using the PDB2PQR and PROPKA tools and noncovalent bond determination algorithms. A user-friendly interface allows visualizing pH-sensitive residues, affected salt bridges, hydrogen bonds and aromatic (pi–pi and cation–pi) interactions. patcHwork can be used to identify patches, a new concept we propose of pH-sensitive residues in close proximity on the protein, which may have a major impact on function. We demonstrate the attractiveness of patcHwork studying experimentally investigated pH-sensitive proteins (https://patchwork.biologie.uni-freiburg.de/).

Reprogramming the Transition States to Enhance C–N Cleavage Efficiency of Rhodococcus opacus l-Amino Acid Oxidase

“…The initial coordinates of the Ro LAAO variants were constructed using AlphaFold2 (See ″ ab initio modeling of RoLAAO with AlphaFold2 ″ for details in Supporting Information ). The protonation states of the charged residues were determined at a constant pH of 8.0, based on pK a calculations using the PROPKA program 56 and considering the local hydrogen bonding network. In the Ro LAAO wild-type/variant M3 models, residues His13, 296, 400, and 436 were set as HIE, and residues His76, 89, 385, 460, 469, 481, and 483 were set as HID.…”

Green Slow-Release Coated Nitrogen Fertilizer with High Water Retention Based on Core–Shell-Structured Self-Healing Supramolecular Networks

“…For dynamic covalent bonding, the recombination of cross-linked networks requires high temperature (up to 120 °C or higher) conditions. In contrast, noncovalent bonding methods, including hydrogen bonds, , ionic interactions, metal–ligand interactions, and host–guest recognition, show better commercialization prospects in SHR construction because their healing behaviors can be easily triggered by light, magnetic, pH, and thermal . Unfortunately, the mechanical properties of rubber synthesized by this method are poor, while the incorporation of common fillers such as graphene, carbon nanotubes, and cellulose nanocrystals would cause a strong interaction between the fillers and the rubber chain, thus reducing the self-healing ability of rubber materials .…”

Machine learning prediction of enzyme optimum pH

“…As a result, deep learning models hold great promise for high-throughput screening of vast metagenomic datasets, such as MGnify, 29 which currently comprises nearly 3 billion non-redundant sequences, as well as large mutant protein libraries to identify proteins with desirable properties. Researchers have employed computational approaches to investigate protein pH relationships with biophysical methods, [30][31][32] and to predict enzyme pHopt using traditional machine learning models with limited datasets (fewer than 500 proteins). 7,[33][34][35][36][37][38][39] Despite these efforts, the adoption of large-scale deep learning to capitalize on recent breakthroughs in the field, remains limited.…”