Accurate and Rigorous Prediction of the Changes in Protein Free Energies in a Large-Scale Mutation Scan

Prediction of mutation-induced protein stability changes based on the geometric representations learned by a self-supervised method

BACKGROUND

Thermostability is a fundamental property of proteins to maintain their biological functions. Predicting protein stability changes upon mutation is important for our understanding protein structure-function relationship, and is also of great interest in protein engineering and pharmaceutical design.

RESULTS

Here we present mutDDG-SSM, a deep learning-based framework that uses the geometric representations encoded in protein structure to predict the mutation-induced protein stability changes. mutDDG-SSM consists of two parts: a graph attention network-based protein structural feature extractor that is trained with a self-supervised learning scheme using large-scale high-resolution protein structures, and an eXtreme Gradient Boosting model-based stability change predictor with an advantage of alleviating overfitting problem. The performance of mutDDG-SSM was tested on several widely-used independent datasets. Then, myoglobin and p53 were used as case studies to illustrate the effectiveness of the model in predicting protein stability changes upon mutations. Our results show that mutDDG-SSM achieved high performance in estimating the effects of mutations on protein stability. In addition, mutDDG-SSM exhibited good unbiasedness, where the prediction accuracy on the inverse mutations is as well as that on the direct mutations.

CONCLUSION

Meaningful features can be extracted from our pre-trained model to build downstream tasks and our model may serve as a valuable tool for protein engineering and drug design.

Titin domains with reduced core hydrophobicity cause dilated cardiomyopathy

The underlying genetic defect in most cases of dilated cardiomyopathy (DCM), a common inherited heart disease, remains unknown. Intriguingly, many patients carry single missense variants of uncertain pathogenicity targeting the giant protein titin, a fundamental sarcomere component. To explore the deleterious potential of these variants, we first solved the wild-type and mutant crystal structures of I21, the titin domain targeted by pathogenic variant p.C3575S. Although both structures are remarkably similar, the reduced hydrophobicity of deeply buried position 3575 strongly destabilizes the mutant domain, a scenario supported by molecular dynamics simulations and by biochemical assays that show no disulfide involving C3575. Prompted by these observations, we have found that thousands of similar hydrophobicity-reducing variants associate specifically with DCM. Hence, our results imply that titin domain destabilization causes DCM, a conceptual framework that not only informs pathogenicity assessment of gene variants but also points to therapeutic strategies counterbalancing protein destabilization.

Prioritizing Small Sets of Molecules for Synthesis through in-silico Tools: A Comparison of Common Ranking Methods

Prioritizing molecules for synthesis is a key role of computational methods within medicinal chemistry. Multiple tools exist for ranking molecules, from the cheap and popular molecular docking methods to more computationally expensive molecular-dynamics (MD)-based methods. It is often questioned whether the accuracy of the more rigorous methods justifies the higher computational cost and associated calculation time. Here, we compared the performance on ranking the binding of small molecules for seven scoring functions from five docking programs, one end-point method (MM/GBSA), and two MD-based free energy methods (PMX, FEP+). We investigated 16 pharmaceutically relevant targets with a total of 423 known binders. The performance of docking methods for ligand ranking was strongly system dependent. We observed that MD-based methods predominantly outperformed docking algorithms and MM/GBSA calculations. Based on our results, we recommend the application of MD-based free energy methods for prioritization of molecules for synthesis in lead optimization, whenever feasible.

The guidance and adhesion protein FLRT2 dimerizes in cis via dual small-X3-small transmembrane motifs

Fibronectin Leucine-rich Repeat Transmembrane (FLRT 1-3) proteins are a family of broadly expressed single-spanning transmembrane receptors that play key roles in development. Their extracellular domains mediate homotypic cell-cell adhesion and heterotypic protein interactions with other receptors to regulate cell adhesion and guidance. These in trans FLRT interactions determine the formation of signaling complexes of varying complexity and function. Whether FLRTs also interact at the surface of the same cell, in cis, remains unknown. Here, molecular dynamics simulations reveal two dimerization motifs in the FLRT2 transmembrane helix. Single particle tracking experiments show that these Small-X₃-Small motifs synergize with a third dimerization motif encoded in the extracellular domain to permit the cis association and co-diffusion patterns of FLRT2 receptors on cells. These results may point to a competitive switching mechanism between in cis and in trans interactions, which suggests that homotypic FLRT interaction mirrors the functionalities of classic adhesion molecules.

Computational insights into lipid assisted peptide misfolding and aggregation in neurodegeneration

An overview of recent advances in computational investigation of peptide–lipid interactions in neurodegeneration – Alzheimer's, Parkinson's and Huntington's disease.

Molecular basis for the increased affinity of an RNA recognition motif with re-engineered specificity: A molecular dynamics and enhanced sampling simulations study

The RNA recognition motif (RRM) is the most common RNA binding domain across eukaryotic proteins. It is therefore of great value to engineer its specificity to target RNAs of arbitrary sequence. This was recently achieved for the RRM in Rbfox protein, where four mutations R118D, E147R, N151S, and E152T were designed to target the precursor to the oncogenic miRNA 21. Here, we used a variety of molecular dynamics-based approaches to predict specific interactions at the binding interface. Overall, we have run approximately 50 microseconds of enhanced sampling and plain molecular dynamics simulations on the engineered complex as well as on the wild-type Rbfox·pre-miRNA 20b from which the mutated systems were designed. Comparison with the available NMR data on the wild type molecules (protein, RNA, and their complex) served to establish the accuracy of the calculations.

Free energy calculations suggest that further improvements in affinity and selectivity are achieved by the S151T replacement.

RNA is an outstanding target for oncological intervention. Engineering the most common RNA binding motif in human proteins (called RRM) so as to bind to a specific RNA has an enormous pharmacological potential. Yet, it is highly non trivial to design RRM-bearing protein variants with RNA selectivity and affinity sufficiently high for clinical applications. Here we present an extensive molecular simulation study which shed light on the exquisite molecular recognition of the empirically-engineered complex between the RRM-bearing protein Rbfox and its RNA target pre-miR21. The simulations allow predicting a variant, the S151T, which may lead to further enhancement of selectivity and affinity for pre-miR21.

On the Applicability of Force Fields To Study the Aggregation of Amyloidogenic Peptides Using Molecular Dynamics Simulations

Molecular dynamics simulations play an essential role in understanding biomolecular processes such as protein aggregation at temporal and spatial resolutions which are not attainable by experimental methods. For a correct modeling of protein aggregation, force fields must accurately represent molecular interactions. Here, we study the effect of five different force fields on the oligomer formation of Alzheimer's Aβ_16-22 peptide and two of its mutants: Aβ_16-22(F19V,F20V), which does not form fibrils, and Aβ_16-22(F19L) which forms fibrils faster than the wild type. We observe that while oligomer formation kinetics depends strongly on the force field, structural properties, such as the most relevant protein-protein contacts, are similar between them. The oligomer formation kinetics obtained with different force fields differ more from each other than the kinetics between aggregating and nonaggregating peptides simulated with a single force field. We discuss the difficulties in comparing atomistic simulations of amyloid oligomer formation with experimental observables.

Toward Improved Understanding of the Interactions between Poorly Soluble Drugs and Cellulose Nanofibers

Cellulose nanofibers (CNFs) have interesting physicochemical and colloidal properties that have been recently exploited in novel drug-delivery systems for tailored release of poorly soluble drugs. The morphology and release kinetics of such drug-delivery systems heavily relied on the drug-CNF interactions; however, in-depth understanding of the interactions was lacking. Herein, the interactions between a poorly soluble model drug molecule, furosemide, and cationic cellulose nanofibers with two different degrees of substitution are studied by sorption experiments, Fourier transform infrared spectroscopy, and molecular dynamics (MD) simulation. Both MD simulations and experimental results confirmed the spontaneous sorption of drug onto CNF. Simulations further showed that adsorption occurred by the flat aryl ring of furosemide. The spontaneous sorption was commensurate with large entropy gains as a result of release of surface-bound water. Association between furosemide molecules furthermore enabled surface precipitation as indicated by both simulations and experiments. Finally, sorption was also found not to be driven by charge neutralization, between positive CNF surface charges and the furosemide negative charge, so that surface area is the single most important parameter determining the amount of sorbed drug. An optimized CNF-furosemide drug-delivery vehicle thus needs to have a maximized specific surface area irrespective of the surface charge with which it is achieved. The findings also provide important insights into the design principles of CNF-based filters suitable for removal of poorly soluble drugs from wastewater.

Making sense of the past: hyperstability of ancestral thioredoxins explained by free energy simulations

Thioredoxin (Trx), a small and globular protein, is present in all kinds of organisms, from Archea to higher mammals. Throughout evolution, the Trx sequence has undergone subtle modifications to adapt to varying environmental conditions. The high degree of sequence conservation makes Trx very amenable to ancestral protein reconstruction techniques. In this work, we address the study of the structural and energetic determinants of thermostability in E. coli Trx using a dataset of mutations inspired by ancestral reconstruction. We compute, from first principles, the expected contribution of 19 different amino acid substitutions to the stability (ΔΔG) and the melting temperature (ΔT_m) of the protein. We also describe the specific changes in structure and protein dynamics responsible for the stabilizing or destabilizing effects of these mutations. Our results point to local and independent changes for most of the variants. Our predictions are accurate enough to substantiate the proposal of new hypotheses regarding evolutionary relationships between mutations, as in the case of T89R, P68A and G74S or K90L and F102A, and reach beyond the initial set to suggest improved variants, such as K90I or K90Y.

Robust Prediction of Resistance to Trimethoprim in Staphylococcus aureus

The rise of antibiotic resistance threatens modern medicine; to combat it new diagnostic methods are required. Sequencing the whole genome of a pathogen offers the potential to accurately determine which antibiotics will be effective to treat a patient. A key limitation of this approach is that it cannot classify rare or previously unseen mutations. Here we demonstrate that alchemical free energy methods, a well-established class of methods from computational chemistry, can successfully predict whether mutations in Staphylococcus aureus dihydrofolate reductase confer resistance to trimethoprim. We also show that the method is quantitatively accurate by calculating how much the most common resistance-conferring mutation, F99Y, reduces the binding free energy of trimethoprim and comparing predicted and experimentally measured minimum inhibitory concentrations for seven different mutations. Finally, by considering up to 32 free energy calculations for each mutation, we estimate its specificity and sensitivity.